Re: What was a 3314?
edgould1...@comcast.net (Edward Gould) writes: > It addressing had MMBBCCHHR(R?) so I guess you could address it > directly. Anyone remember how to do that? (progr5amming for a 2321 is > a lost art (where is Seymour?). the "BB" was to select the BIN that the magnetic strips were located in. https://www-03.ibm.com/ibm/history/exhibits/storage/storage_2321.html and https://en.wikipedia.org/wiki/IBM_2321_Data_Cell Generically, at the OS level, IBM defined the six bytes as BBCCHH, for Bin, Bin, Cylinder, Cylinder, Head and Head respectively. ... snip ... referenced from: http://www.bitsavers.org/pdf/ibm/360/os/R21.7_Apr73/GC28-6628-9_OS_System_Ctl_Blks_R21.7_Apr73.pdf the bins (cells) rotated under the r/w heads. https://www-03.ibm.com/ibm/history/exhibits/storage/storage_PH2321B.html http://www.computer-history.info/Page4.dir/pages/Photostore.dir/images/Picture.1.jpg as undergraduate, univ. hired me as fulltime support for the production ibm systems. the univ. library got an ONR grant to do online catalog ... and used part of the money to get a 2321 datacell. The project was also selected to be one of the original CICS product betatest sites and I got tasked with debugging/supporting CICS (one of the "bugs" was that CICS original implementation at customer site used specific set of BDAM options which was hardcoded in the source ... and the library had chosen a different set of BDAM options ... it took some dump analysis to discover the issue since it wasn't documented). -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: TCP/IP question on routing
rpomm...@sfgmembers.com (Pommier, Rex) writes: > Sorry if these are silly questions, but my brain is really foggy this > morning. My questions are for validation of what I think would happen > with various iterations of IPCONFIG DATAGRAMFWD. > > Scenario 1, I have a single IP address on my z/OS system running to a > single network segment out an OSA port. In this scenario, DATAGRAMFWD > would have no effect, correct? > > Scenario 2, I have 2 IP addresses, connected via 2 different OSA > adapters to 2 different networks. If I have DATAGRAMFWD, when TCPIP > sees a packet coming in on OSA 1 but it has a destination address of a > device on network 2, TCP/IP will forward the packet out the other OSA, > effectively acting like a router or gateway. If I set it to > NODATAGRAMFWD, if TCP/IP sees the same packet on OSA1, it will ignore > the packet rather than forwarding it. Is this how it actually works? > > What I'm looking at is we have a new machine being installed that I > need to be able to access from 2 different networks at different > times. These two networks need to be isolated from each other and I > don't want the mainframe to start acting as a router, passing packets > from one to the other. Is there some other configuration setting I > need to be aware of, or would the NODATAGRAMFWD be sufficient to > keeping them isolated? > > Yes, the safest solution would be to unplug one of the network cables, > and just have 1 plugged in at a time, but the machine is about 300 > miles from me. modulo bugs in the code. the weekend before 1988 Interop, the floor nets were crashing well into early Monday morning ... before problem was identified. As a result, new requirements were mandated in RFC1122 (RFC1122 & RFC1123 combined are official internet STD3). regarding automagically forwarding packets. ... Any host that forwards datagrams generated by another host is acting as a gateway and MUST also meet the specifications laid out in the gateway requirements RFC [INTRO:2]. An Internet host that includes embedded gateway code MUST have a configuration switch to disable the gateway function, and this switch MUST default to the non-gateway mode. In this mode, a datagram arriving through one interface will not be forwarded to another host or gateway (unless it is source-routed), regardless of whether the host is single- homed or multihomed. The host software MUST NOT automatically move into gateway mode if the host has more than one interface, as the operator of the machine may neither want to provide that service nor be competent to do so. ... snip ... trivia: i had hardware in booth at Interop 88 ... but not in the IBM booth ... which was on the other side of the floor. -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: opinion? Fujitsu USA
elardus.engelbre...@sita.co.za (Elardus Engelbrecht) writes: > I vaguely remember that I worked [indirectly] with them when I started > worked around 1989. > > ICL [from Britain?] and Amdahl [from that wizard Gene Amdahl] were > guzzled up by Fujitsu. Fujitsu was major manufacture and investor in Amdahl (from the start). Not long after Amdahl Co. was formed (early 70s), Gene had presentation at MIT in large, full auditorium ... some of us went over from the IBM Cambridge Science Center. Several in the audience pressed Gene pretty hard about being front for foreign interests (regarding the ties with Fujitsu) Later my boss at IBM was head of the workstation IBU (PC/RT, RS/6000, etc) and had some head-to-head with some senior executives and left, eventually forming HAL ... early 64bit RISC (also backed by Fujitsu). There was joke that they was so much traffic back between silicon valley and japan that it justified the non-stop San Jose/Narita flight (and some companies had permanently reserved 1st class seats). Topic Drift Warning. One of the issues was corporate mandate that RS/6000 had to use PS2 microchannel cards (and not do their own). As mentioned in recent "Tech News 1964" posting, the communication group was fiercely fighting off client/server and distributed computing ... and PS2 microchannel cards had minimal throughput and performance. we had snide comments that if RS/6000 was restricted to to only using PS2 cards, it wouldn't have any better throughput/performance than PS2. For instance, the PS2 (32bit) microchannel 16mbit T/R card had lower throughput than the PC/RT (16bit) atbus 4mbit T/R card (a PC/RT server with 4mbit T/R card would have higher throughput than RS/6000 with 16mbit T/R card ... i.e. the workstation IBU had done their own 4mbit T/R card but were prevented from doing their own 16mbit T/R card. http://manana.garlic.com/~lynn/subtopic.html#801 The communication group had design point for 16mbit T/R card of 300+ stations sharing common LAN doing terminal emulation. The major justification for having done T/R was many large mainframe customers were running into bldg. weight loading restrictions with the massive amount of 3270 coax cables. T/R LAN enormously reduced the wiring needed, the aggregate LAN bandwidth increased the number of stations per LAN ... but terminal emulation didn't require significant per card throughput. In the late 80s, the communication group 16mbit T/R card was something like $899/card ... but there were $69 10mbit Ethernet cards running over CAT5 wiring, that had higher per card throughput than the 16mbit T/R card. The new Almaden Research bldg had extensive CAT5 wiring assuming 16mbit T/R, and found that not only the $69 ethernet cards had higher per card throughput ... but running 10mbit ethernet had higher aggregate LAN throughput and lower latency than 16mbit T/R. The communication group publications comparing 16mbit T/R with ethernet apparently used original 3mbit ethernet that did have listen before transmit protocol. ACM SIGOPS publication had paper that did detailed study of typical Ethernet configurations and found effective aggregate throughput was around 9mbits/mbit ... but running low-level device driver code that constantly transmitted minimum sized ethernet packets on all stations ... the effective aggregate LAN throughput dropped off to 8mbit/sec. The lower aggregate 16mbit T/R LAN throughput and higher latency was attributed to the token passing processing latency. http://manana.garlic.com/~lynn/subnetwork.html#terminal disclaimer: my wife is one of the inventors on one of the original IBM token-passing patents. -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Fwd: Tech News 1964
peter.far...@broadridge.com (Farley, Peter x23353) writes: > IMHO part of what is vanishing mainframe clients is IBM's failure > several decades back to continue to support universities with > discounted hardware and software. Lack of mainframe availability at > university level has translated into current managements with no > exposure and no desire to learn the advantages (TCO, security, etc.) > of mainframes. Not the whole reason, but a significant contributor. The gov. legal action resulted in number of IBM responses ... 23jun1969 unbundling announcement that included starting to charge for (application) software; it also saw IBM pull back from the enormous grants and discounts it gave academic institutions... some past posts http://manana.garlic.com/~lynn/submain.html#unbundle IBM did come back in the early 80s with the academic business unit (ACIS) ... it was putting several hundred million into universities (but lots of it would go into non-mainframe technologies). IBM also sponsored the university BITNET (where this ibm-main mailing list originated). some past posts http://manana.garlic.com/~lynn/subnetwork.html#bitnet it used technology similar to IBM internal network ... some past posts (larger than arpanet/internet from just about the beginning until sometime mid-80s) http://manana.garlic.com/~lynn/subnetwork.html#internalnet that originated at the IBM cambridge science center ... some past posts http://manana.garlic.com/~lynn/subtopic.html#545tech this was non-SNA (and not communication group technology) ... at about the same time in the late 80s when the communication group was forcing the internal network into moving to SNA ... BITNET moved to tcp/ip (which would have been much better for internal network also, rather than SNA). I've told the story several times about senior disk engineer at late 80s, annual communication group world-wide internal conference. His talk was supposedly on 3174 performance but he opened with the statement that the communication group was going to be responsible for the demise of the disk division. The issue was that the communication group had strategic "ownership" of everything that cross the datacenter walls and was fiercely fighting off client/server and distributed computing trying to preserve their dumb terminal paradigm install base. The disk division was seeing applications fleeing the datacenter to more distributed computing friendly platforms with drop in disk sales. The disk division had come up with a number of solutions to try and correct the situation, but were constantly being vetoed by the communication group. some past posts http://manana.garlic.com/~lynn/subnetwork.html#terminal Somewhat as work-around to the communication group opposition, the disk division VP of software was investing in open system and distributed computing technology ... the POSIX support in MVS ... as well as startups that built mainframe based distributed computing hardware and software solutions. trivia: the original mainframe tcp/ip product was implemented in vs/pascal (which had none of the buffer length related exploits that have been epidemic in C-language implementations) ... however for various reasons it only got about 44kbytes/sec using full 3090 processor. I did the enhancements to support RFC1044 and in some tuning tests at Cray Research between Cray and 4341, got sustained channel speeed throughput using only a modest amount of 4341 processor ... possibly 500 times improvement in bytes moved per instruction executed. some past posts http://manana.garlic.com/~lynn/subnetwork.html#1044 -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Microprocessor Optimization Primer
note that test was on both 360/67 and 360/65 machines and was atomic. I've commented before about charlie invented compare (chosen because CAS are his initials) while doing fine-grain multiprocessor locking working on CP67 (360/67 precursor to vm370) at the science center. http://manana.garlic.com/~lynn/subtopic.html#545tech and http://manana.garlic.com/~lynn/subtopic.html#smp then we attempted to get it added to 370 architecture. initially was rebuffed because the POK favorite son operating system people said that test was more than adequate for multiprocessor support (serializing critical code sections). The 370 architecture owners said that to get it justified would require additional uses, not just multiprocessor serialization. Thus was invented the multiprogramming/multithreading examples (used whether or not running on multiprocessor machine) that still are shown in the principles of operation. The problem in a multithreaded application is it is enabled for interrupts and can loose control in a locked/critical section. Compare is used for doing an atomic operation directly not needing to lock a critical section. This was especially leveraged by large multiprogramming/multithreading DBMS avoiding needing to make kernel calls for lots of serialization ... and by the 80s lots of other platforms (especially those supporting high-throughput DBMS) were including compare (or instructions with similar semantics). I first saw transactional memory on 801/risc in the late 70s. They demonstrated that they could do transactional type operations on applications that weren't originally coded for transactions. 801/risc ROMP (research/office products) that started out going to be a displaywriter followon. When the displaywriter followon was canceled, they looked around and decided to retarget it to the workstation market. They hired the company that had done the UNIX port to IBM/PC for PC/IX to do one for romp. This was eventually released as PC/RT and AIX. The followon to ROMP was RIOS (rs/6000) and they used the transactional memory to implement JFS ... journalling the UNIX filesystem metadata changes ... with a claim that it was more efficient that directly implementing journalling calls in the filesystem. However, Palo Alto then did a portable JFS that used explicit journaling calls ... and demonstrated on RS/6000 that it was much faster than the transaction memory implemention. http://manana.garlic.com/~lynn/subtopic.html#801 Note that s/370 had very strong (multiprocessor) memory consistency and cost huge amount in performance. Two processor multiprocessor machines slowed each processor clock cycle by 10% to accommodate cross-cache protocol chatter ... and this overhead went up non-linearly. Later IBM mainframe was running cache machine cycle at much higher rate than the processor machine cycle. In the late 80s, I was asked to participate in the standardization (started by LLNL) of what quickly became fibre-channel standard (on which they eventually built the heavy-weight FICON protocol that drastically reduces the native throughput) http://www.garlic.com/~lynn/submisc.html#ficon I was also asked to participate in the standardization of scalable coherant interface (started by people at SLAC ... a large VM370 mainframe installation at the time and host of the monthly IBM BAYBUNCH user group meetings). SCI was defined for both I/O operations as well as multiprocessor shared memory operation. The standard SCI memory concistency defined 64-port memory bus ... that relaxed memory concistency (compared to IBM mainframe) and allowed for lot larger mainframe configuration.s Sequent, Data General, Silicon Graphics, and at least Convex built multiprocessor products. Sequent & Data General took standard i486 four processor board that shared cache and built interface to SCI ... being able to get 64 4-processor boards in configurations (256-way processor shared memory configuration). Convex took standard HP/SNAKE (risc) two processor board that shared cache and built interface to SCI ... being able to get 64 2-processor boards in configuration. As an aside, much later IBM buys Sequent and shuts it down. Note both FCS and SCI started out with fiber that supported concurrent transfers in both direction. SCI https://en.wikipedia.org/wiki/Scalable_Coherent_Interface is part of what evolves into infiniband https://en.wikipedia.org/wiki/InfiniBand other trivia ... in the mid-70s I was involved in project that defined a 16-way shared memory multiprocessor. Lots of people thought it was really fantastic ... and we got some of the 3033 processor engineers to work on it in their part time (lot more interesting than mapping 168 logic to 20% faster chips). Then somebody tells the head of POK that it could be decades before the POK favorite son operating system could effectively support 16-way (it was 2000 before 16-way shipped) and we got invited to never visit POK again (and the 3033 processor engineers were
Re: CeBIT and mainframes
dcrayf...@gmail.com (David Crayford) writes: > Emulex sells an HBA that handles over 1M IOPS on a single port. IIRC, > x86 Xeon class servers have something called DDIO which facilitates > writes directly to processor cache. > It's not too dissimilar to offloading I/O to SAPs. I've got old > colleagues that work on distributed now and they are of the opinion > that I/O bandwidth is not an issue on x86 systems, > but it's not exactly commodity hardware. They're all hooked up using > 16Gbs fiber connected to a SAN using PCIe, the same as z Systems. > > I would question the RAS capabilities rather than I/O. Last published mainframe I/O I've seen was peak I/O benchmark for z196 which got 2M IOPS using 104 FICON (running over 104 fibre-channel). Also that all 14 SAPs would run 100% busy getting 2.2M SSCHs/sec but recommendation was keeping SAPs to 75% or 1.5M SSCHs/sec. About the same time of the z196 peak I/O benchmark there was fibre-channel announced for e5-2600 blade claiming over million IOPS, two such fibre-channel getting more throughput than 104 FICON (running over 104 fibre-channel) ... aka FICON is enormously heavy-weight protocol that drastically cuts the native throughput of fibre-channel. disclaimer: 1980 I was asked to do the support for channel extender for STL (now IBM Silicon Valley Lab), they were moving 300 people from the IMS group to offsite bldg. with access back to the STL datacenter; they had tried remote 3270 but found the human factors intolerable. The channel extender support put channel attached 3270 controllers out at the offsite bldg ... and resulted in response indistinguishable from channel attach 3270 controllers within the STL bldg. The vendor they tried to get approval from IBM to release the support, but there was a group in POK that was playing with some serial stuff and they got it blocked because they were afraid it might interfer with getting their stuff released. In 1988, I'm asked to help standardize some serial stuff that LLNL was playing with which quickly becomes fibre channel standard ... one of the issues is that protocol latency effects increases with increase in bandwidth ... so that it becomes apparent at relatively short distances. One of the features with the 1980 work is that it localized the enormous IBM channel protocol latency at the offsite bldg and then used much more efficient protocol the longer distance. For fibre-channel used the much more efficient protocol for everything. In 1990, the POK group finally get their stuff release as ESCON when it is already obsolete. Then some POK engineers become involved with fibre channel standard and define a protocol that enormously cuts the native throughput ... that is eventually released as FICON. Note that the more recent zHPF/TCW work for FICON looks a little more like the work that I had done back in 1980. Besides the peak I/O benchmark FICON throughput issue (compared to native fibre channel issue) there is also the overhead of CKD simulation. There hasn't been any real CKD disks built for decades, current CKD disks are all simulation on industry standard commodity disks. Other tivia, when I moved to San Jose Research in the 70s, they let me wander around. At the time the disk engineering lab (bldg 14) and disk product test lab (bldg 15) they were running pre-scheduled standalone mainframe around the clock, 7x24. At one point they had tried to us MVS for concurrent testing, but found that MVS had 15mins MTBF in that environment. I offerred to rewrite I/O supervisor that made it bullet proof and never fail ... being able to do ondemand, anytime concurrent testing, greatly improving productivity. I happened to mention that MVS 15min MTBF in an internal-only report on the work ... which brings down the wrath of the MVS group on my head (not that it was untrue, but that it exposed the information to the rest of the company). When they found that they couldn't get me fired, they then were to make sure they made my career as unpleasant as possible (blocking promotions and awards whenever they could). z900, 16 processors, 2.5BIPS (156MIPS/proc), Dec2000 z990, 32 processors, 9BIPS, (281MIPS/proc), 2003 z9, 54 processors, 18BIPS (333MIPS/proc), July2005 z10, 64 processors, 30BIPS (469MIPS/proc), Feb2008 z196, 80 processors, 50BIPS (625MIPS/proc), Jul2010 EC12, 101 processors, 75BIPS (743MIPS/proc), Aug2012 z13 published refs is 30% move throughput than EC12 (or about 100BIPS) with 40% more processors ... or about 710MIPS/proc z196 era e5-2600v1 blade rated at 400-500+BIPS depending on model, e5-2600v4 blades are three-four times that, around 1.5TIPS (1500BIPS). i.e. since the start of the century, commodity processors have increased their processing power significantly more aggresively than mainframe. They have also come to dominate the wafer-chip manufacturing technology ... and essentially mainframe chips have converged to use the same technology (in much the same way mainframe has converged to use
Re: CeBIT and mainframes
dave.g4...@gmail.com (Dave Wade) writes: > In fact its a bit like SVC's in VM/370. The code which handles them is > very different to that in the OS world, but the code still runs there was joke about the time MVS came out with 8mbyte kernel image in every virtual address space ... that the 32kbyte os/360 system services simulation in VM/CMS was a lot more efficient than the 8mbyte os/360 system services simulation in MVS. -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Introducing the New z13s: Tim's Hardware Highlights
edgould1...@comcast.net (Ed Gould) writes: > Remember the *OLD* days there was a 16MB max on (even) an MP? Never > mind the cost of $10K per meg (if memory serves me on a 168). > Yes the newer machines have more memory but in reality you really > don't get all that more functionality, and yes there are bells and > whistles for the z genation. Significant MVS bloat by 3033 was causing a number of problems ... real storage requirements was banging hard at the 16mbyte limit. 16bit 370 PTE was 12bit (4kbyte) page number, 2defined bits and 2undefined/unused. They took 2undefined/unused bits then used them to prefix the (real) page number ... allowing 14bit page number or up to 64mbytes of real pages ... allowing lots of application virtual pages to reside above the 16mbyte line. os/360 significant pointer passing API paradigm was making 16mbyte virtual address space limit a problem. Transition from SVS to MVS gave each application its own 16mbyte virtual address space ... but pointer passing API paradigm required 8mbyte image of the MVS kernel in each application virtual address space. Then because subsystems services were in their own virtual address space, pointer passing API required 1mbyte CSA (in each virtual address space) for passing parameters. CSA size requirements were proportional to subsystems and applications ... for large 3033s was 5-6mbytes and threatening to become 8mbytes (leaving none for applications). Subset of "access registers" was then retrofitted to 3033 as dual-address mode (allowing subsystems to access application virtual address space w/o needing CSA). problem was that 4341 clusters had more processing power than 3033, more aggregate memory and I/O throughput, much lower cost and significantly less physical and environmental footprint. Folklore is that head of POK felt so threatened that corporate was convinced to cut allocation of critical 4341 manufacturing component in half. 4341 had significant improvement price/performance as well as physical and environmental footprint resulted in corporations ordering hundreds at a time for placing out in departmental areas ... sort of the leading edge of distributed computing tsunami. Before 4341s shipped, I got roped into benchmarking engineering 4341 for national labs for big compute farm ... sort of the leading edge of the coming supercomputer paradigm internet+distributed computing+compute farms ... evolves into cloud with hundreds of thousands of systems and millions of processors in each cloud megadatacenter (system costs have dropped to such a level that power are starting to dominate cloud costs). old email about air force data systems coming out to talk about 20 4341s, spring of 1979 (they had a few mainframes in their datacenter), but by the time they got around to caming out fall of 1979, it had jumped to 210 4341s. http://www.garlic.com/~lynn/2001m.html#email790404 and http://www.garlic.com/~lynn/2001m.html#email790404b other 4341 related email http://www.garlic.com/~lynn/lhwemail.html#4341 -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: History of Computing 1944 and the evolution to the System/360
charl...@mcn.org (Charles Mills) writes: > My *recollection* is that the S/360 30 came with up to 48K, or 64K by RPQ. I > could be off, but 1MB sounds incredibly high to me. ga24-3231-7, 360-30 functional characteristics pg14 (from bitsavers) c308kbytes d30 16kbytes dc30 24kbytes e30 32kbytes f30 64kbytes univ had 709/1401 and was sold 360/67 replacement (for tss/360) ... pending delivery of 360/67, transition replaced 1401 with 64kbytes 360/30 ... gave univ. chance to get acquated with 360 ... but 360/30 could be also be run in 1401 hardware emulation mode. tss/360 never quite came to production fruition ... so 360/67 ran most of the time as 360/65 with os/360. IBM offered 2361 large capacity storage https://en.wikipedia.org/wiki/IBM_2361_Large_Capacity_Storage models came in 1mbyte and 2mbyte for models 50, 65, and 75. i also remember ampex (and other vendors) offering LCS up to 8mbytes, also additional memory for 30s & 40s. search engine turns up other vendors offering addon semiconductor/monolithic memory for 360s in the 70s with larger sizes at cheaper prices. -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: ASCII vs. EBCDIC (was Re: On sort options ...)
000a2a8c2020-dmarc-requ...@listserv.ua.edu (Tom Marchant) writes: > ASCII was seriously considered for the initial System/360 > design. Amdahl, Blaauw and Brooks published an article in the IBM > Journal in April, 1964, titled "Architecture of the System/360" in > which many of the design trade-offs were described. One place where > the article can be found is > http://web.ece.ucdavis.edu/~vojin/CLASSES/EEC272/S2005/Papers/IBM360-Amda= hl_april64.pdf > > > ASCII vs BCD codes. The selection of the 8-bit character size in 1961 > proved wise by 1963, when the American Standards Association adopted a > 7-bit standard character code for information interchange > (ASCII). This 7-bit code is now under final consideration by the > International Standards Organization for adoption as an international > standards recommendation. The question became =E2=80=9CWhy not adopt ASCI= I as > the only internal code for System/360?=E2=80=99 > > The reasons against such exclusive adoption was the widespread use of > the BCD code derived from and easily translated to the IBM card > code. To facilitate use of both codes, the central processing units > are designed with a high degree of code independence, with generalized > code translation facilities, and with program-selectable BCD or ASCII > modes for code-dependent instructions. Nevertheless, a choice had to > be made for the code-sensitive I/O devices and for programming > support, and the solution was to offer both codes, as a user > option. Systems with either option will, of course, easily read or > write I/O media with the other code. > IBMer Bob Bemer ... "father of ASCII" ... EBCDIC and the P-Bit (The Biggest Computer Goof Ever). http://www.bobbemer.com/P-BIT.HTM Who Goofed? The culprit was T. Vincent Learson. The only thing for his defense is that he had no idea of what he had done. It was when he was an IBM Vice President, prior to tenure as Chairman of the Board, those lofty positions where you believe that, if you order it done, it actually will be done. I've mentioned this fiasco elsewhere. Here are some direct extracts: ... snip, see reference for a whole lot more ... have been having problems with outgoing email this weekend, apologize if multiple copies show up -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Query: Will modern z/OS and z/VM classes suffice for MVS and VM/370
ri...@livingcomputermuseum.org (Rich Alderson) writes: > We are currently in the process of restoring a 4341 to operating > condition. We have just last week corrected a fault in the power > system, and are able to power the system up and IML it from floppy. > > We are now deciding what operating system to run on the restored > system. Most likely, we will run VM/370, but possibly we will run an > MVS guest as well. I used to be an MVS systems programmer, but that > was more than 30 years ago, and even the rust has eroded away. > > I would like to brush up on operations and systems programming, which > would be much simpler if a modern z/OS and/or z/VM course would > suffice for the older operating systems. Have the operator commands > and programming utilities changed radically since 1984 (JES2, CMS)? > > Please feel free to reply privately if you wish to tell me how foolish this > sounds. > > Thanks, > Rich Alderson Hercules comes with 4341 era vm370 https://en.wikipedia.org/wiki/Hercules_%28emulator%29 vast majority of 4341s were shipped with FBA disks ... you would need some sort of CKD disks in order to bring up MVS. huge percentage of 4341s went out into departmental areas with 3370 FBA disks, sort of leading edge of distributed computing tsunami ... not requiring datacenter provisioning. -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: AW: Re: You thought IEFBR14 was bad? Try GNU's /bin/true code
dlc@gmail.com (David L. Craig) writes: > Does anyone else (Google doesn't) remember the ELHO acronym? > > Equal- mask '8' > Low - mask '4' > High - mask '2' > Overflow - mask '1' > > Back in the days of no extended mnemonic opcodes it was > quite the assembler programming aid. I was involved in doing some of the original relational/sql "System/R", some past posts http://www.garlic.com/~lynn/submain.html#systemr where SQL does 2-value, true/false logic. It created big problem for unknowns and/or null values. Old discussion that null values in SQL tended to produce the opposite of the expected results ... making it all the more dangerous. about the same time, I was brought in to do some of a different kind of relational implementation ... which had interface language that directly supported unknowns/null with 3-value logic. old post in DBMS group discussing the dangers of unknowns/nulls in SQL and how it was handled in 3-value logic http://www.garlic.com/~lynn/2003g.html#40 How to cope with missing values - NULLS? http://www.garlic.com/~lynn/2003g.html#41 How to cope with missing values - NULLS? Earlier in the first part of the 70s, I had written a PLI program to ingest assembler listing ... analyze the statements, creating higher level representation ... code paths, logic processes, "dead" code, possible register use before set, etc ... and generate representation using a psuedo-pascal statements. some highly optimized cp67 kernel tmade liberal use of 3&4-way logic with branch conditions and would appear very straight forward ... the psuedo-pascal true/false if/then/else type logic could look very convoluted with nesting that could go 15-levels deep. iefbr14 https://en.wikipedia.org/wiki/IEFBR14 got fixed with SR, then with eyecatcher name, also a couple other nits, "IEFBR14" on "END" statement, and "RENT" & "REUSE" in linkedit http://hercules390.996247.n3.nabble.com/IBM-program-naming-question-td27674.html from above: Two more ... First, because the RENT linkedit option wasn't specified. Second, because the REUSE linkedit option wasn't specified. ... snip ... -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Ancient History (OS's) - was : IBM Destination z ...
000433f07816-dmarc-requ...@listserv.ua.edu (Paul Gilmartin) writes: > (Topic drift on recreation) I found a fun Mandelbrot set viewer at: other IBM Mandelbrot drift ... In the 80s, Mandelbrot resigned from IBM Research in protest over the elimination of research. https://en.wikipedia.org/wiki/Benoit_Mandelbrot Mandelbrot left IBM in 1987, after 35 years and 12 days, when IBM decided to end pure research in his division.[20] He joined the Department of Mathematics at Yale, and obtained his first tenured post in 1999, at the age of 75.[21] At the time of his retirement in 2005, he was Sterling Professor of Mathematical Sciences. ... snip ... -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: IBM Destination z - What the Heck Is JCL and Why Does It Look So Funny?
t...@tombrennansoftware.com (Tom Brennan) writes: > Yep - I'm hoping they'll like the batch facilities in MVS which in my > opinion are far beyond unix. This might be a spot where a history > lesson is needed, but I wasn't around in the early days: > > From what I've read, MVS started with nothing but batch jobs and later > grew into online systems. So TSO is just another batch job that > happens to communicate with a terminal. On the unix side though, it > seems they started with online terminals first, so a batch > (background) job was later created as a terminal session with no > terminal. I've pointed that out before ... CTSS was conversational online default from the start, then some of the people went to the 5th flr and did Multics (and folklore is that some of the Bell Labs people went back home and did a simplified Multics, calling it unix) ... and other of the people went to the science center on the 4th flr and did CP/40-CMS (making hardware modifications to 360/40 to support virtual memory), which morphs into CP/67-CMS when standard 360/67 with virtual memory standard comes available precursor to VM/370-CMS (cms originally stood for "cambridge monitor system" ... is renamed "converstational monitor system" for vm/370). some cambridge science center posts http://www.garlic.com/~lynn/subtopic.html#545tech I've periodically mentioned Kildall working with CP/67-CMS at NPG school, before doing CP/M, which then morphs at Seattle Computing, and leads to ms/dos. os/360 assumed batch ... and had to provide increasing about of contingency handling capability ... while conversational started out assuming responsible human was there to handle the contigency cases. we were working with director of NSF and were suppose to get $20M to tie together the the NSF supercomputer centers. Then congress cuts the budget, some other things happen, and then they release an RFP ... but internal politics prevent us bidding on the RFP (director of NSF tries to help by writing the corporation a letter, but that just makes internal politics worse). As regional networks tie into the centers, it morphs into NSFNET backbone, precursor to modern internet some old email http://www.garlic.com/~lynn/lhwemail.html#nsfnet We do get TCP/IP product for mainframe but there are quite a few issues ... getting 44kbytes/sec using near full 3090 processor. I do the enhancements to support RFC1044 and in some tuning tests at Cray Research get full sustained channel speed throughput between 4341 and a Cray, using only a modest amount of 4341 (possibly 500 times improvement in bytes moved per instruction executed). some past posts http://www.garlic.com/~lynn/subtopic.html#rfc1044 Late 90s, a senior disk engineer gets a talk scheduled at annual, world-wide, internal communication group conference, supposedly on 3174 performance ... but opens the talk with statement that the communication group was going to be responsible for the demise of the disk division. The issue was that the communication group had strangle hold on datacenters with corporate strategic ownership of everything crossing the datacenter walls, and were fiercely fighting off distributed computing and client/server (trying to preserve their dumb terminal paradigm and install base). The disk division was starting to see data fleeing the datacenter to more distributed computing friendly platforms with drop in disk sales. The disk division had come up with a number of solutions to reverse the process, but were constantly being vetoed by the communication group. some past posts http://www.garlic.com/~lynn/subnetwork.html#terminal In the late 60s, increasing number of cp/67-cms customers were extending to 7x24 availability (including some number of commercial online service bureaus). One of the issues in the 60s mainframes were rented and initially, it was hard to promote offshit use enough to recover system costs. There was a lot of work done to reduce system costs (especially offshift). Part of system rental costs were based on the system meter that ran whenever the processor or any channel was running. All processor and channel activity had to be quiet for at least 400ms before system meter would stop. Special terminal CCWs were created to allow channel to stop ... but immediately start ondemand when characters were coming in ... some of the stuff sort of analogous being done for on-demand cloud computing (trivia long after mainframes had converted from rental to sales, MVS still had timer task that woke up every 400ms ... making sure that system meter never stopped). Other stuff to further minimize offshift costs was eliminating operator requirements. Another early CP/67 enhancement in the 60s was automatic re-ipl after failure (system come up and available w/o needing any human intervention). In the early 70s, as environments became more complex, increasing amount of CP/67 services were provided by "service virtual machines" (analogous to demons ... the current
Re: IBM Destination z - What the Heck Is JCL and Why Does It Look So Funny?
harris...@gmail.com (Graham Harris) writes: > Doesn't deadline scheduling count? as undergraduate in the 60s, I did dynamic adaptive resource management that was picked up and shipped in CP/67 (customers periodically referred to as fairshare scheduler or wheeler scheduler because default policy was fairshare). http://www.garlic.com/~lynn/subtopic.html#fairshare in the morph from CP/67 to VM/370 there was a lot of things dropped and simplified ... including all the scheduling stuff. At the science center during the FS period, I continued to work on 360 & 370 stuff ... even periodically ridiculing the FS activity. http://www.garlic.com/~lynn/submain.html#futuresys with death of FS, there was mad rush to get stuff back into 370 product pipelines which contributed to decision to release some amount of the stuff I had been doing. Some of it was shipped in standard release. Note that earlier in the 23Jun1969 unbundling announcement, ibm started to charge for se services, maintenance, (application) software ... but made the case that kernel software should still be free. http://www.garlic.com/~lynn/submain.html#unbundle During the FS period, the lack of 370 products is credited with given clone processors a market foothold. So part of resuming 370 efforts, the decision was made to also transition to start charging for all kernel software (likely motivated by 370 clone makers getting market foothold). The decision was made to make the scheduling work a guinee pig as separate charged-for kernal product (I had to spend a lot of time with lawyers and business people about kernel charging policies). After the transition was complete to charging for all kernel software in the 80s ... the next step was the OCO-wars (aka only shipping object code). As part of the product review process somebody in Armonk said he wouldn't approve it unless it had customer setable parameters because everybody knew that the state of the art was setable performance parameters (MVS would have this huge array of setable paremters ... there would be lots of SHARE presentations about various tests of random walks of all the setable parameters with various workloads). I tried to explain to him what dynamic adaptive management met ... but eventually had to implement some customer setable parameters. However, there was a joke that I took from operations research and "degrees of freedom". The range of values for the manually setable parameters was less than what the dynamic adaptive calculations could do ... so effectively the dynamic adaptive calculations could compensate for human selected values. -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Ancient History (OS's) - was : IBM Destination z - What the Heck Is JCL and Why Does It Look So Funny?
linda.lst...@comcast.net (Linda) writes: > I had an Apple ][ with an acoustic coupler. It auto dialed over a > regular telco dial tone line using a program loaded from a cassette > player, or if one could afford it, from an early floppy drive. The > college I went to had a Univac 90/70d. The were 4 student dialup > numbers. I could get into one of those much like the scene from War > Games. It was fun. TYMSHARE made their CMS-based online computer conferencing available free to SHARE as VMSHARE starting in Aug1976 ... archives: http://vm.marist.edu/~vmshare In the 70s, I started trying to get IBM to let me put all the VMSHARE files up on internal systems ... including the world-wide sales support HONE system. One of the biggest battles I had with IBM was the lawyers were afraid that customer information would contaminate IBM employees. My brother was Apple regional marketing rep at the time (largest physical region in CONUS) and I started trying to get him to setup up an apple that would do terminal emulation for copying all the files down from TYMSHARE ... he never quite got around to doing it ... although over the years ... when he would come into town for business meetings I would get invited to dinners ... and even got to argue with the MAC developers about design (before MAC was announced). I eventually had to resort to getting montly tapes mailed from TYMSHARE ... that dumped all VMSHARE files (later added all PCSHARE files). misc. old email http://www.garlic.com/~lynn/lhwemail.html#vmshare -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Ancient History (OS's) - was : IBM Destination z - What the Heck Is JCL and Why Does It Look So Funny?
bles...@ofiglobal.com (Lester, Bob) writes: > Yeah. Worst mistake Gary Kindall ever made. Just think, if he'd hadn't > "blown off" IBM, I'd be cursing his memory (he's deceased) instead of > Bill Gates. Or maybe not, I ran CP/M-80 back in the day. I really > enjoyed it. But, then, I enjoyed everything more back then. > everything was bright, shiny, and new before ms/dos http://en.wikipedia.org/wiki/MS-DOS there was seattle computer http://en.wikipedia.org/wiki/Seattle_Computer_Products before seattle computer there was cp/m, http://en.wikipedia.org/wiki/CP/M before cp/m, kildall worked with cp67/cms (precursor to vm370) at npg http://en.wikipedia.org/wiki/Naval_Postgraduate_School other trivia ... after 64, commodore did amiga ... which ran ARexx https://en.wikipedia.org/wiki/ARexx ARexx is an implementation of the REXX language for the Amiga, written in 1987 by William S. Hawes, with a number of Amiga-specific features beyond standard REXX facilities. Like most REXX implementations, ARexx is an interpreted language. Programs written for ARexx are called "scripts", or "macros"; several programs offer the ability to run ARexx scripts in their main interface as macros. ... snip ... more trivia ... acorn group in Boca kept claiming that they wouldn't going to do any software and an IBM group was formed in silicon valley to write software for acorn. Then at some point the Boca group changed their mind and wanted responsibility for all software ... if necessary contracting with outside groups (some viewed as eliminating internal competition). some past mentioning acorn http://www.garlic.com/~lynn/2002g.html#79 Coulda, Woulda, Shoudda moments? http://www.garlic.com/~lynn/2003d.html#19 PC history, was PDP10 and RISC http://www.garlic.com/~lynn/2005q.html#24 What ever happened to Tandem and NonStop OS ? http://www.garlic.com/~lynn/2005r.html#8 Intel strikes back with a parallel x86 design http://www.garlic.com/~lynn/2006y.html#29 "The Elements of Programming Style" http://www.garlic.com/~lynn/2007e.html#5 Is computer history taugh now? reference https://en.wikipedia.org/wiki/IBM_Personal_Computer#Project_Chess -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: 3270 based ATMs
other trivia: A Brief History of the ATM http://www.theatlantic.com/technology/archive/2015/03/a-brief-history-of-the-atm/388547/ The company seemed poised to overwhelm its competitors until executives decided to deploy a new model "the IBM 4732 family" which were incompatible with previous models, including the already-successful and widely deployed IBM 3624. ... IBM's move soured banks, inadvertently, opening the ATM market to new cashpoint manufacturers. Eventually, IBM abandoned payment-technology systems entirely. ... IBM's returns fell short of its expectations, in part due to the growth in local processing architectures, which had invalidated IBM's strategy to link ATMs to its expensive mainframes. ... snip ... There are articles about OS/2 lingering on in ATM market long after it had disappeared elsewhere The ticking time bomb inside your bank ATM http://www.fiercefinanceit.com/story/ticking-time-bomb-inside-your-bank-atm/2013-07-31 -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: 3270 based ATMs
tro...@gmail.com (Rick Troth) writes: > I searched before asking, but didn't find anything close. > Anyone know how many 3270 based ATMs are in operation? > Anyone know where I can find tech pubs for such? 3624 designed at los gatos lab (disclaimer at one time, I had wing of offices and labs there ... not involved in 3624, but heard stories, at one time los gatos labs was considered one of the most scenic in ibm ... since plowed under and now housing development) https://en.wikipedia.org/wiki/IBM_3624 the guy managing magstripe standards was also there https://en.wikipedia.org/wiki/Magnetic_stripe_card#Further_developments_and_encoding_standards when I was co-author of financial industry standard that was piloted by nacha for ATM ... this reference gone 404 ... but lives on at wayback machine ... pilot results entry 23July2001: http://web.archive.org/web/20070706004855/http://internetcouncil.nacha.org/News/news.html all the ATM network stuff was Tandem ... with special crypto hardwware. -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Lineage of TPF
000248cce9f3-dmarc-requ...@listserv.ua.edu (Ed Finnell) writes: > As Lynn mentioned there were hardware mods for ACP/TPF to the 3081, 3083 > and 3090's. They were given new numbers 9081,9083 and of course 9190? I guess > shorter path lengths and such but couldn't find any details after a short > search. besides the 3830 disk controller RPQ ... the 3083 was 3081 with one of the processors removed (at the time, acp/tpf didn't have tightly-coupled multiprocessor support) that still wasn't competitive ... so there was 3083 with specialized channel microcode operation tailored to ACP/TPF operation. I'm not familiar something similar for 3090. as mentioned 3081 technology wasn't competitive with clones: http://www.jfsowa.com/computer/memo125.htm initial 3081D per processor throughput was suppose to be faster than 3033 ... but many benchmarks have it about 20% slower. 3081K doubled the cache and per processor was suppose to improve to 50% faster than 3033 ... but many benchmarks were same as 3033. IBM 2-way multiprocessor technology from the period slowed the processor clock down by 10% to handle cross-cache activity. Going from 3081K to 3083K increased processor clock by nearly 15% (no multiprocessor clock slow-down) ... 3083 mostly done because all ACP/TPF customers might migrate to clone makers (since ACP/TPF didn't have multiprocessor support). Faster clock and tweaks for 3083jx got it up to 16% faster than 3081K (or supposedly almost 80% faster than 3033). 9083 had different I/O microcode load to bias for the typical higher channel i/o loads by ACP/TPF. It is possible that they may have done something similar for 3090, but I don't recollect any details. -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Where do you place VSE?
g...@gabegold.com (Gabe Goldberg) writes: > One response cited Wikipedia entry. ALSO good timing; I'm ALSO writing > article on VSE community. As you'd expect, the VSE list has had a lot > to say -- positive, negative, and informative. OS/360 for a time PCP, MFT, and MVT ... but didn't work well in the smallest real memory configurations ... giving rise to DOS/360. OS/360 somewhat becomes split between MFT customers (usually mid-size memory configurations) and MVT (largest memory configurations). Move to virtual memory, DOS/360 morphs into DOS/VS (singe virtual address space, MFT morphs into VS1 (single virtual address space) and MVT morphs into VS2 (initially with single virtual address space, aka SVS much like VS1 ... and eventually MVS with multiple virtual address spaces). During the Future System period, 370 efforts are being killed off (FS was different than 360/370 and was going to completely replace it). With the demise of FS, there is mad rush to get products back into the 370 pipeline. POK kicks off 3033 (168 logic mapped to 20% faster chips) and 3081 & MVS/XA in parallel (and convinces corporate to kill off vm370 product and move all the people to POK to work for MVS/XA; Endicott eventually acquires the vm370 product mission, but has to recreate a development group from scratch). While POK is doing "XA" architecture ... highly tailored to MVS ... Endicott kicks off the "E" architecture ... which in large part is moving the single virtual address space into microcode and new instructions that enable/disable virtual page for specific real page. Internally the 4331 is called E3 and 4341 is called E4. DOS/VS becomes DOS/VSE. In part because large percentage of 4300 machines are run with vm/370 ... they are actually run in 370 mode ... supporting 370 multiple virtual address spaces. os/vs1 https://en.wikipedia.org/wiki/OS/VS1 above slightly garbled since the migration aid was primarily motivated for helping move mvs/370 to mvs/xa VSE https://en.wikipedia.org/wiki/VSE_%28operating_system%29 The migration aid originally was only going to be used for internal mvs/xa development and never released to customers and so paid little attention to general function and performance. There then is internal politics ... an internal datacenter added full XA support to VM370 with full function/performance. POK wants corporate to support a massive new staff for the migration aid to try and upgrade it to the feature and performance of standard vm370 (with XA added). POK wins. -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Lineage of TPF
g...@gabegold.com (Gabe Goldberg) writes: > Indeed. Then a couple people responded. Good timing; I'm writing > article on TPF for Destination z or IBM Systems Magazine (I forget > where it'll be published). IBM TPFers have been very helpful and I'm > contacting TPF users group: http://www.tpfug.org/ . I didn't post here > because well, I just didn't, but I should have. Better late than > never: I'm interested in TPF insights, experiences, etc. > > Be brief, this won't be an epic article, though there might be > follow-on pieces. Please copy me directly so replies aren't buried in > the list digest. > > Thanks... > > Rick Troth said on IBM-Main: Lineage of TPF would also be interesting. Before Jim left for tandem (earlier post about RDBMS, System/R, DB2) ... he was looking for real live DBMS locking statistics for profiling System/R (RDBMS) performance. This included data from ACP 3830 controller RPQ ... logical/symbolic locks implemented in the 3830 controller (much more efficient than device reserve/release) ... minor note IBM wanted to depreciate the ACP RPQ because corporate strategy was to push "string switches" ... which allowed two different controllers to get to the same device (and bypass "locks" in the other controller). old email refs: http://www.garlic.com/~lynn/2008i.html#email800325 The customer statistics in above ... was just before the looming 3081 "crisis" (while ACP/TPF had loosely-coupled cluster support, it didn't have tightly-coupled, SMP support). Note above mentions two controlers (with string-switch) ... but it is same system having access to both controllers for redundancy. As aside, US HONE datacenters were consolidated in Silicon Valley in the mid-70s (HONE was the world-wide, online sales support system). By the late 70s, the US HONE system had the largest "single-system image" loosely-coupled configuration in the world (with load-balancing and recovery across all systems in the complex) ... and required string-switch with pairs of controllers each with multiple channel connections in order for all the SMP (multiprocessor) systems in the complex to fully access the large DASD farm. Rather than locking (device reserve/release) for the necessary operations, it used a special CCW sequence (when needed) that emulated the compare-and-swap instruction semantics. past posts mentioning charlie invented compare-and-swap while doing fine-grain multiprocessing locking on CP67 at the science center ... past posts http://www.garlic.com/~lynn/subtopic.html#smp and http://www.garlic.com/~lynn/subtopic.html#545tech trivia: when facebook moved to silicon valley it was to a new bldg. next door to the old HONE datacenter. misc. past posts mentioning HONE http://www.garlic.com/~lynn/subtopic.html#hone Of course the (HONE cluster) "single system image" support wasn't made available to customers until 30yrs later (late last decade). -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Lineage of TPF
jo.skip.robin...@att.net (Skip Robinson) writes: > I had a brief and bemusing encounter with TPF around 1990. My > employer, Security Pacific Bank, was acquired by (the old SF-based) > Bank of America, which was then under the tutelage of an ex CEO of > American Airlines. He believed that TPF was the answer to all > important IT questions. In particular, he engineered a project to > manage the Bank's ATMs with TPF, perhaps the only time/place that TPF > was charged with that responsibility--absolutely critical for a major > financial institution. It apparently worked pretty well. My mainframe > buddies there admired TPF for its lightning quick recovery--a > blessing, they said, because it crashed a lot. ;-) There was fantastic SE on financial institution account in LA ... he wrote ATM cash machine support in VM370 that he showed had higher throughput on 370/158 than TPF ran on 370/168. His trick was significant better disk arm scheduling (than TPF) ... had patterns of ATM useage and record layout ... and do things like delaying transaction somewhat proportional to the record location distance from current arm position and the probability another transaction would come in needing a record closer to the current arm position. -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Lineage of TPF
sas...@sas.com (Don Poitras) writes: > TPF ran lots of ATM networks. I worked at First Interstate in 1988 > working on a project to convert from TPF to MVS. And certainly any > bank that does VISA authorization at their ATMs still to this day use > TPF because that's what VISA uses. (credit) card associations started out as non-profits for brand marketing (getting merchants to accept brands) and network interconnect between merchant acquiring and customer issuing financial institutions (at one time 30,000 institutions). Interchange rules (amount charge merchants for credit card transaction, had pieces for the acquiring and issuing financial institutions as well as piece for the association network (and other surcharges based on things like fraud history & probability) VISA installation for ACP for its card association network transaction processing was possibly big part of changing name from ACP to TPF. The card associations were making so much money off their netework interchange transaction charges (for card association networks) that they changed to profit and spun off in IPOs. Around the turn of the century ... because of bank consolidation and outsourcing, 90% of credit card transactions were handled by six datacenters that had direct connections and no longer needed the card association networks. There was then big legal battle between card associations and the six processors (who felt they no longer had to share interchange fees with the card associations ... since they were no longer using their networks). ATM/Debit networks were primarily Tandem (even though the backends might be IBM mainframes). Tandem had also acquired major ATM machine crypto hardware vendors. Long ago I got brought in as consultant into a small client/server that wanted to do financial transactions on their server, they had also invented this technology they called SSL they wanted to use, the result is now frequently called "electronic commerce". Somewhat for having done "electronic commerce", in the mid-90s I got asked to work in the X9A10 financial standard working group that had been given the requirement to preserve the integrity of the financial infrastructure for all retail payments. We did detailed end-to-end threat analysis for nearly all kinds of retail payments (credit, debit, ACH, wire-transfer, face-to-face, point-of-sale, internet, etc). The result was a standard that eliminated most of the current kinds of fraud ... the downside was interchange fees have been heavily prorated based on fraud rates ... with an enormous profit component ... actually eliminating the fraud enormously impacts those calculations (and profit). NACHA Internet Council did debit pilot with support in the Tandem network processors ... results published 23July2001, gone 404 but lives on at the wayback machine http://web.archive.org/web/20070706004855/http://internetcouncil.nacha.org/News/news.html Compaq/Tandem had previously sponsored large workshop for me Jan1999 on the financial protocol standards ... old long-winded post by somebody at the workshop http://www.garlic.com/~lynn/aepay3.htm#riskm The CEO of one of the companies that we had been working with and at the meeting, had been the head of POK mainframes in a prior life. tandem ref (which includes reference to Jim Gray, which I worked with at IBM and left for Tandem ... but by the mid-90s was at Microsoft https://en.wikipedia.org/wiki/Tandem_Computers and https://en.wikipedia.org/wiki/Jim_Gray_%28computer_scientist%29 -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Lineage of TPF
edgould1...@comcast.net (Ed Gould) writes: > I was not on the the team (next cubicle over). I was somewhat involved > in the precursor(?) of Mastercard called Town & Country. This was in > Chicago. The OS that Mastercard was was written was DOS (I *THINK* it > was on a 360/30) and to some extent MFT (350/50) (this goes back 40 or > so years so please forgive the memory errors). I do not have anything > to add to the mastercard/ and the VISA (I just do not remember what > the name was). I will take as face value about the battle, although I > do remember it somewhat. mastercard had huge number of series/1 in their network interfacing between acquirers and issuers. around the turn of decade ... the populace was moving from credit to debit ... and the card associations introduced "signature debit" at point-of-sale ... that ran through the credit networks (and had credit level fraud and the much higher credit/fraud interchange fees ... rather than pin-debit through the debits network that had much lower fees as well as card associations not getting anything). national retailer association then had anti-trust legal action against the card associations for forcing debit point-of-sale transactions as "signature debit" (with significantly higher fees) ... and won huge settlement. card associations then came up with "cash back" as an alternative ... where the cash back interchange rate that merchants pay is significantly higher than the "cash back" that consumers actually see (this is eventually going to replace the enormous amounts they make off fraud surchange fees ... when the get around to deploying more secure technology) -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Lineage of TPF
other trivia ... IBM had bought complex that had been originally built in Purchase for new Nestle hdqtrs (before Nestle ever moved in). In the 90s, during the IBM troubles ... the new CEO was looking to raise cash and was selling off real estate (even at well below market and sometimes even below originally paid) ... and sold the Purchase bldg to MasterCard for its new hdqtrs. We had a meeting there (to discuss online banking) shortly after MasterCard moved in, they said that they paid more to have all the door hardware handles to be replaced ... than they paid IBM for the complex (something like 1% of the original Nestle building cost). -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Man Versus System
john.archie.mck...@gmail.com (John McKown) writes: > Descended from ACP (Airline Control Program). > https://en.wikipedia.org/wiki/IBM_Airline_Control_Program > > I worked at Braniff Airways before it went under. The reservation system > ran ACP on a 2 Meg 3033. The thing would IPL in about 5 seconds. The ACP > systems people were a bit strange. They had the source and modified it. I > remember the CE complaining that the ACP attached tapes (3420s) would just > die with "no warnings at all" whereas the MVT (yes MVT on a 3033) and, a > bit later MVS and VM would show temp errors. The ACP people then told the > CE that they had removed all logging of temporary errors to speed up > processing. Not just on the tapes, but on the 3344 disks as well. IIRC, the > 3344s on ACP actually used "software duplexing" for reliability. there was big problem with 3081 ... which originally was going to be multiprocessor only ... and ACP/TPF didn't have multiprocessor support (they were afraid that the whole market would move to clone processors which were building newer single processor machines). An an interim they shipped some number of releases of VM370 with very unnatural things done to it specifically for running ACP/TPF on multiprocessors (but degraded performance for all other customers). Eventually they shipped 3083 ... which was a 3081 box with one of the processors removed (minor trivia, simplest would be to remove the 2nd processor which was in the middle of the box ... but that would have made the box dangerously top-heavy, they had to rewrire "processor 0" to processor in the middle and remove the processor at the top of the box). other issues with 308x ... highlighting that it was warmed over FS technology http://www.jfsowa.com/computer/memo125.htm later in the 80s, my wife did temporary stint as chief architect for Amedeus (euro res system based on old eastern "system one") ... the communication group got her replaced because she backed x.25 (instead of sna/vtam) ... it didn't do them much good because amedeus went with x.25 anyway. later in the mid-90s, we were asked to look at re-engineering some of the largest airline res system in the world ... starting with ROUTES (about 25% of total mainframe processing load) addressing the ten impossible things that they couldn't do. I went away and two months later came back with totally different ROUTES implementation that ran about hundred times faster and did all ten impossible things ... including ten RS/6000 990s being able to handle every ROUTES transaction for every airline in the world. The issue was much of ACP/TPF implementation was dictated by technology trade-offs made in the 60s ... it was possible to start from scratch 30yrs later and make totally different trade-offs (and a decade later, cellphone processors had processing capacity of those ten 990s). It was fun because they provided me with tape of the full OAG ... including record for every scheduled airline flt in the world. -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Compile error
jo.skip.robin...@att.net (Skip Robinson) writes: > The name 'DB2' seems to have followed the 1980s tradition of what I call > 'name bloat', the practice of inflating a moniker in one way or another to > make a product look more mature or more elegant. The paragon in my mind was > dBASE II from Ashton-Tate. There never was a plain old dBASE. The roman > numeral was added from the get-go to make the product seem new and improved. > Moreover, there was never an 'Ashton'. That name was invented because, gosh > darn it, it sounded good hyphenated with Tate, a real person. > > Before DB2 there was precedent for name bloat within IBM. There never was a > plain old 'JES'. The product emerged from the cocoon as JES2. There had been > a predecessor product called 'HASP', which may or may not have been an > acronym for Houston Automatic Spooling Priority, but the name 'J-E-S' was > born complete with suffix. > > Meanwhile there did emerge a 'JES3', but it was not an evolutionary > descendant of JES2. Both products have coexisted, albeit uneasily, for > decades. We used to imagine a JES5 or JES6 (depending on one's arithmetic > proclivity) that would somehow combine the best features of both products, > but it's almost certainly DOA. Likewise, the prospects for a 'DB3' are as > dim as a distant star. note that VS1 had JES1 (Job Entry Subsystem 1) https://en.wikipedia.org/wiki/OS/VS1 The official names were OS/VS1 and OS/VS2 ... so JES2 originally may have originally been to designate it was for OS/VS2. Long ago and far away, my wife was in the GBURG JES group and was part of the catchers for ASP turning into JES3. She was then co-author of JESUS (JES UNIFIED SYSTEM) document ... which merged the features in JES2 and JES3 that respective customers couldn't live w/o ... for various reasons never saw the light of day. A Fascinating History of JES2 http://www.share.org/p/bl/et/blogid=9=238 For the truth we must go back to the mid 1960's. IBM's OS/360 was in trouble. The spooling (wonder where that name came from) support was slow and the overhead was high. Many programming groups independently attacked the problem. ASP, loosely based upon the tightly coupled IBM 7090/7094 DCS, held the lead in the OS/360 spooling sweepstakes. ASP's need for at least two CPU's fit well with IBM Marketing's plans for the System/360. Meanwhile, a group of IBM SE's, located in Houston, developed a different product of which they were justifiably proud. They wanted to popularize it, as they correctly suspected it would be the balm for OS/360 users, increasing the usability and popularity of the operating system, and, not incidentally, furthering their careers. All they needed was the right name! A name which was easy to remember, a name which would draw attention to their product, and a name to distract from the ASP publicity. That name was Half-ASP, or HASP. Naturally, if HASP and ASP were products of two different companies, the FTC would have stepped in to stop such a predatory product name. Regulatory action was prevented, however, because IBM is "one big happy family", believed by many to be larger than the Government. ... snip ... of course officially, the "H" stands for "Houston" https://en.wikipedia.org/wiki/Houston_Automatic_Spooling_Priority then my wife was con'ed into going to POK to be responsible for loosely-coupled architecture ... where she "peer-coupled shared data architecture" ... which saw very little uptake (except for IMS hot-standby) until SYSPLEX & Parallel SYSPLEX ... contributing to her not remaining long in POK (along with the ongoing periodic battles with the communication group trying to force her into using SNA/VTAM for loosely-coupled operation). some past posts http://www.garlic.com/~lynn/submain.html#shareddata as undergraduate in the 60s, I got to make a lot of HASP modifications (I had also been hired fulltime by the university to be responsible for production mainframe systems) ... including implementing terminal support and conversational editor in HASP for a form of CRJE. https://en.wikipedia.org/wiki/Remote_job_entry DB2 may have been because some had hopes that the official new DBMS "EAGLE" might still be able to rise from its ashes ... or it was to designate the OS/VS2 (aka MVS) version of System/R as opposed to the earlier SQL/DS version of System/R (that ran on VM370, VS1, DOS/VSE). trivia: one of the problems with the System/R tech transfer to Endicott for SQL/DS ... was that several enhancements to vm370 had been made to make System/R much more efficient. For various reasons, the Endicott people didn't want to make SQL/DS release dependent on getting enhancements into VM370 ... and so that had to be dropped. -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message:
Re: Compile error
other trivia from ibm jargon: MVM - n. Multiple Virtual Memory. The original name for MVS (q.v.), which fell foul of the fashion of changing memory to storage. MVS - n. Multiple Virtual Storage, an alternate name for OS/VS2 (Release 2), and hence a direct descendent of OS. OS/VS2 (Release 1) was in fact the last release of OS MVT, to which paging had been added; it was known by some as SVS (Single Virtual Storage). MVS is one of the big two operating systems for System/370 computers (the other being VM (q.v.)). n. Man Versus System. ... snip ... as part of the "Man Versus System" theme ... it had become significantly much easier to work out lots of computer concepts and design on vm370/cms ... and then later port the implementation to MVS ... than trying to start on an MVS base. some time ago, I got a request about the history of adding virtual memory to all 370s ... old post with exchange from IBMer involved (who recently passed) with references of os/v2, future systems, hasp/asp, etc http://www.garlic.com/~lynn/2011d.html#73 other parts of the thread http://www.garlic.com/~lynn/2011d.html#71 http://www.garlic.com/~lynn/2011d.html#72 http://www.garlic.com/~lynn/2011d.html#74 -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Compile error
hal9...@panix.com (Robert A. Rosenberg) writes: > And then there was Star Wars (AKA: A New Hope [which was added when > the film was rereleased as part of the release of The Empire Strikes > Back]) which opened with a crawl saying Episode 4". That was just > because they were emulating the old serials where each segment was a > numbered Chapter with its own title (which often reflected the > cliffhanger being resolved or the plot point of that chapter). co-worker at IBM would talk about Lucas attending San Jose Astronomy club meetings and bringing draft outlines for all 8 episodes (for members to review) More recent interviews with Lucas says that the first episode he chose to do, was the one most likely for getting funding. -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Compile error
thomas.sa...@fiserv.com (Savor, Thomas , Alpharetta) writes: > Management System or DBMS in 1983 when IBM >released DB2 on its MVS > mainframe platform." -- Wikipedia, citing an IBM manual as authority. > > All these years, I've have only known of DB2. The name seems to have stuck. > > Was there ever a DB1 ?? > Will there ever be a DB3 ?? The original sql/relational implementation was at SJR (bldg. 28 on main plant site, using modified vm/370 on 370/145), System/R. History/Reunion: http://www.mcjones.org/System_R/ wiki https://en.wikipedia.org/wiki/IBM_System_R and another history http://www.theregister.co.uk/2013/11/20/ibm_system_r_making_relational_really_real/ and http://www.cs.ubc.ca/~rap/teaching/504/2010/readings/history-of-system-r.pdf The official new DBMS project was EAGLE with the corporation focused on EAGLE it was possible to get System/R out the door as SQL/DS (under the radar). When EAGLE imploded, there was a request about how fast would it take to port System/R to MVS ... eventually released as DB2 (originally for analytical & decision support *only*). past posts mentioning System/R http://www.garlic.com/~lynn/submain.html#systemr also referenced here http://www.mcjones.org/System_R/citations.html The Birth of SQL http://www.mcjones.org/System_R/SQL_Reunion_95/sqlr95-The.html Some discussion of EAGLE and then DB2 http://www.mcjones.org/System_R/SQL_Reunion_95/sqlr95-DB2.html I periodically reference this post about Jan1992 meeting in Ellison's conference room http://www.garlic.com/~lynn/95.html#13 one of the people in the meeting would tell how he was responsible for the majority of the tech transfer into the Santa Teresa Lab (now silicon valley lab) for DB2. Jim Gray departs for Tandem palming off some number of things on me ... old email ref: http://www.garlic.com/~lynn/2007.html#email801016 Eventually IBM Toronto starts RDBMS for IBM/PC ... implemented in C .. which is made available on other platforms and is also called DB2 ... even though it is totally different code base from the mainframe implementation. SQL/DS is also eventually renamed DB2 https://en.wikipedia.org/wiki/IBM_SQL/DS -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Fibre Chanel Vs FICON
Kevin Bowlingwrites: > I'm shortly going to be the new owner of a z800 at home. Looking > forward to booting and playing with this bistro, what kind of disk array > do I need? Is fibre channel storage enough, or is FICON extra special > at the protocol level? Is there any way to network boot/emulate storage > or will I be looking for FICON arrays next? there are two issues ... one is the FICON protoool running over fibre-channel standard ... some past posts http://www.garlic.com/~lynn/submisc.html#ficon and controller emulation of CKD on industry standard fixed-block disks (there haven't been any real CKD manufactured for decades). http://www.garlic.com/~lynn/submain.html#dasd there have been various past discussions about IBM charging/justifying a significant $$/mbyte premium for that emulation trivia Build Your Own Fibre Channel SAN For Less Than $1000 - Part 1 http://www.smallnetbuilder.com/nas/nas-howto/31485-build-your-own-fibre-channel-san-for-less-than-1000-part-1 -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: History question - In what year did IBM first release its DF/DSS backup & restore product?
ibmm...@computersupervisoryservices.com (Stephen Mednick) writes: > Looking to find the answer to the question "in which year did IBM release > its DF/DSS backup & restore product. some trivia from the web https://www.ibm.com/developerworks/community/blogs/InsideSystemStorage/entry/ibm_storwize_product_name_decoder_ring1?lang=en In my post January 2009 post [Congratulations to Ken on your QCC Milestone], I mentioned that my colleague Ken Hannigan worked on an internal project initially called "Workstation Data Save Facility" (WDSF) which was changed to "Data Facility Distributed Storage Manager" (DFDSM), then renamed to "ADSTAR Distributed Storage Manager" (ADSM), and finally renamed to the name it has today: IBM Tivoli Storage Manager (TSM). ... snip ... Note: I had originally developed CMSBACK in the late 70s that was used at a number of internal sites (including the online world-wide, sales support HONE). It went through a number of internal releases at San Jose Research ... which morphed into Almaden research when research moved up the hill in the mid-80s. Ability to back up from distributed systems was added and then released to customers as workstation datasave facility. some past email http://www.garlic.com/~lynn/lhwemail.html#cmsback It was picked up by the storage division ... and renamed ADSM ... adstar storage manager ... storage division had been renamed adstar as part of reorganizing the company into the 13 baby blues in preparation for breaking up the company (then the board brought in a new CEO to reverse the breakup and resurrect the company). The company acquired Tivoli (started by a couple former IBMers that had been at the rs/6000 workstation group in Austin) and ADSM was moved to Tivoli morphing into TSM. We spent some amount of time consulting for the ADSTAR VP of software on number of items ... not just ADSM, he also was behind the original MVS/USS development ... and provided funding for some number of non-IBM storage related startups. I've mentioned several times: A senior disk engineer got a talk scheduled at communication group world-wide internal annual conference supposedly on 3174 performance ... however he opened the talk with the statement that the communication group was going to be responsible for the demise of the disk division. The issue was that the communication group had strangle hold on datacenters with its strategic ownership of everything that crossed the datacenter wall and was fighting off distributed computing and client/server trying to preserve its (emulated) dumb terminal paradigm and install base. The disk division was seeing data fleeing to more distributed computing friendly platforms with drop in disk sales. The disk division had come up with a number of solutions to reverse the trend, but they were constantly vetoed by the communication group. ... As attempted work-around (to the communication group) the VP of software would fund non-IBM efforts to provide mainframe distributed computing and part of what he had us do was try and keep track of some number of these activities. -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Is there a source for detailed, instruction-level performance info?
shmuel+ibm-m...@patriot.net (Shmuel Metz , Seymour J.) writes: > We ran more than that, plus TSO, on a 2 MiB machine. IBM executives were looking at 370/165 ... where typical customer had 1mbyte ... in part because 165 real memory was very expensive ... and typical regions were such that they only got four in 1mbytes (after system real storage requirement) later, newer memory for 370/168 was less expensive ... and started to see four mbytes as much more common ... aka four mbytes as 370/165 would have met that typical MVT customer could have gotten 16 regions ... w/o having to resort to virtual memory ... but the decision had already been made. basical initial transtion os/mvt to os/vs2 svs was MVT laid out in single 16mbyte virtual address space ... and little bit of code to build the segement/page tables and handle page faults. The biggest code hit was adding channel program translation in EXCP ... code initially copied from CP67 CCWTRANS channel program translation. prior reference/discssion to justification for 370 virutal memory http://www.garlic.com/~lynn/2011d.html#73 Multiple Virtual Memory later transition to os/vs2 MVS with multiple virtual address spaces ... had other problems. The os/360 MVT heritage was heavily based on pointer passing API paradigm ... and with move to MVS. The first pointer passing ... was put an 8mbyte image of the MVT kernel into every application virtual address space ... leaving only 8mbytes (out of 16) for application use. Then because of subsystems were now in their own (different) virtual address space ... needed a way for passing parameters & data back and forth between applications and subsystems using pointer passing API. The result was "common segment" ... a one mbyte area that also appeared in every virtual address space which could be used to pass arguments/data back between applications and subsystems (leaving only 7mbytes for applications). The next issue was demand for common segment was somewhat proportional to number of concurrent applications and subsystems ... so the common segment area became common system area (CSA) as requirements exceeded 1mbytes. Into the 3033 area, larger operations were pushing CSA to 4&5 mbytes and threatening to push it to 8mbytes ... leaving no space at all for application (of course with MVS kernel at 8mbytes and CSA at 8mbytes, there wouldn't be any left for applications ... which drops the demand for CSA to zero). Part of the solution to address the OS/360 MVT pointer passing API problem was included in the original XA architecture (later referred to 811 ... because documents dated Nov1978) were access registers ... and ability to address/access multiple address spaces. To try and alleviate the CSA explosion in 3033 time-frame ... a subset of access registers was retrofitted to 3033 as dual-address space mode ... but it provide only limited help since it still required updating all the subsystems to support dual-address space mode (instead of CSA). trivia: the person responsible for retrofitting dual-access to 3033 ... later leaves IBM for another vendor and later shows up as one of the people behind HP Snake and later Itanium. -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Is there a source for detailed, instruction-level performance info?
mike.a.sch...@gmail.com (Mike Schwab) writes: > If branch predicting is a big hang up, the obvious solution is to > start processing all possible outcomes then keep the one that is > actually taken. I. E. B OUTCOME(R15) where R15 is a return code of > 0,4,8,12,16. aka, speculative execution ... instructions executed on path ... that is not actually taken ... are not committed https://en.wikipedia.org/wiki/Speculative_execution and https://en.wikipedia.org/wiki/Speculative_execution#Eager_execution Eager execution is a form of speculative execution where both sides of the conditional branch are executed; however, the results are committed only if the predicate is true. With unlimited resources, eager execution (also known as oracle execution) would in theory provide the same performance as perfect branch prediction. With limited resources eager execution should be employed carefully since the number of resources needed grows exponentially with each level of branches executed eagerly.[7] ... snip ... https://en.wikipedia.org/wiki/Eager_evaluation -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Is there a source for detailed, instruction-level performance info?
rpin...@netscape.com (Richard Pinion) writes: > Don't use zoned decimal for subscripts or counters, rather use indexes > for subscripts and binary for counter type variables. And when using > conditional branching, try to code so as to make the branch the > exception rather than the rule. For large table lookups, use a binary > search as opposed to a sequential search. > > These simple coding techniques can also reduce CPU time. in late 70s we would have friday nights after work ... and discuss a number of things ... along the lines of what came up in tandem memos ... aka I was blamed for online computer conferencing on the internal network (larger than arpanet/internet from just about the beginning until sometime mid-80s) in the late 70s and early 80. folklore is that when the corporate executive committee were told about online computer conferencing (and the internal network), 5of6 wanted to fire me. from IBMJARGON: [Tandem Memos] n. Something constructive but hard to control; a fresh of breath air (sic). "That's another Tandem Memos." A phrase to worry middle management. It refers to the computer-based conference (widely distributed in 1981) in which many technical personnel expressed dissatisfaction with the tools available to them at that time, and also constructively criticized the way products were [are] developed. The memos are required reading for anyone with a serious interest in quality products. If you have not seen the memos, try reading the November 1981 Datamation summary. ... snip ... one of the issues was that the majority of the people inside the company didn't actually use computers ... and we thot things would be be improved if the people in the company actually had personal experience using computers, especially managers and executives. So we eventually came up with the idea of online telephone books ... of (nearly) everybody in the corporation ... especially if lookup elapsed time was less than look up of paper telephone book. avg binary search of 256k is 18 ... aka 2*18. Also important was there were nearly 64 entries in physical block ... so binary search to the correct physical block is 12 reads (i.e. 64 is 2**6, 18-6=12). However, it is fairly easy to calculate the name letter frequency ... so instead of doing binary search, do radix search (based on letter frequency) and can get within the correct physical block within 1-3 physical reads (instead of 12). We also got fancy doing first two letter frequency and partially adjusting 2nd probe, based on how accurate the first probe was. In any case, binary search for totally unknown distribution characteristics. So one friday night, we established the criteria, to design, implement, test and deploy the lookup program had to take less than a person week of effort ... and less than another person week to design, implement, test and deploy the process for collecting, formating and distributing the online telephone books. trivia ... long ago and far away ... a couple people I had worked with at Oracle (when I was at IBM and working on cluster scaleup for HA/CMP), had left and were at small client/server responsible for something called commerce server. After cluster scaleup was transferred, announced as IBM supercomputer, and we were told we couldn't work on anything with more than four processors ... we decide to leave. We are then brought in as consultants at this small client/server startup because they want to do payment transactions on the server, the startup had also invented this technology called SSL they want to use, the result is now frequently called "electronic commerce". TCP/IP protocol has session termination process that includes something called FINWAIT list. At the time, session termination was relative infrequent process and common TCP/IP implementations used a sequential search of the FINWAIT list (assuming that there would be few or none entries on the list). HTTP (& HTTPS) implementation chose to use TCP ... even tho it is a datagram protocol rather than a session protocol. There was period in the early/mid 90s as web use was scaling up where webservers saturated spending 90-95% of cpu time doing FINWAIT list searches before the various implementations were upgraded to do significantly more efficient management of FINWAIT (session termination) process. -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Is there a source for detailed, instruction-level performance info?
charl...@mcn.org (Charles Mills) writes: > Not so simple anymore. > > "How long does a store halfword take?" used to be a question that had an > answer. It no longer does. > > My working rule of thumb (admittedly grossly oversimplified) is > "instructions take no time, storage references take forever." I have heard > it said that storage is the new DASD. This is true so much that the z13 > processors implement a kind of "internal multiprogramming" so that one CPU > internal thread can do something useful while another thread is waiting for > a storage reference. > > Here is an example of how complex it is. I am responsible for an "event" or > transaction driven program. I of course have test programs that will run > events through the subject software. How many microseconds does each event > consume? One surprising factor is how fast do you push the events through. > If I max out the speed of event generation (as opposed to say, one event > tenth of a second) then on a real-world shared Z the microseconds of CPU per > event falls in HALF! Same exact sequence of instructions -- half the CPU > time! Why? My presumption is that because if the program is running flat out > it "owns" the caches and there is much less processor "wait" (for > instruction and data fetch, not ECB type wait) time. so such accounting measuring CPU time (elapsed instruction time) is analogous to early accounting which measured by elapsed wall clock time. cache miss/memory access latency ... when measured in count of processor cycles is comparable to 60s disk access when measured in in count of 60s processor cycles. There is lot of analogy between page thrashing when overcommitting real memory and cache misses. This is old account of motivation behind moving 370 to all virtual memory. The issue was that as processors got faster, they spent more and more time waiting for disk. To keep the processors busy, required increasing levels of multiprogramming to overlap execution with waiting on disk. At the time, MVT storage allocation was so bad that a region sizes needed to be four times larger than actually used. As a result, a typical 1mbyte 370/165 would only have four regions. Going to virtual memory, it would be possible to run 16 regions in a typical 1mbyte 370/165 with little or no paging ... significantly increasing aggregate throughput. http://www.garlic.com/~lynn/2011d.html#73 Multiple Virtual Memory risc has been doing cache miss compensation for decades, out-of-order execution, branch prediction, speculative execution, hyperthreading ... can be viewed as hardware analogy to 60s multitasking ... given the processor something else to do while waiting for cache miss. Decade or more ago, some of the other non-risc chips started moving to hardware layer that translated instructions into risc micro-ops for scheduling and execution ... largely mitigating performance difference between those CISC architectures and RISC. IBM documentation claimed that half the per processor improvement from z10->z196 was the introduction of many of the features that have been common in risc implementation for decades ... with further refinement in ec12 and z13. z10, 64processors, aggregate 30BIPS or 496MIPS/proc z196, 80processors, aggregate 50BIPS or 625MIPS/proc EC12, 101 processor, aggregate 75BIPS or 743MIPS/proc however, z13 claims 30% more throughput than EC12 with 40% more processors ... which would make it 700MIPS/processor by comparison z10 era E5-2600v1 blade was about 500 BIPS, 16 processors or 31BIPS/proce. E5-2600v4 blade is pushing 2000BIPS, 36 processors or 50BIPS/proc. as an aside, 370/195 pipeline was doing out-of-order execution ... but didn't do branch proediction or speculative execution ... and conditional branch would drain the pipeline. careful coding could keep the execution units busy getting 10MIPS ... but normal codes typically ran around 5MIPS (because of conditional branches). I got sucked into helping with hyperthreading 370/195 (which never shipped), it would simulate two processors with two instructions streams, sets of registers, etc ... assuming two instruction streams, each running at 5MIPS would then keep all execution units running at 10MIPS. from account of shutdown of ACS-360 http://people.cs.clemson.edu/~mark/acs_end.html Sidebar: Multithreading In summer 1968, Ed Sussenguth investigated making the ACS-360 into a multithreaded design by adding a second instruction counter and a second set of registers to the simulator. Instructions were tagged with an additional "red/blue" bit to designate the instruction stream and register set; and, as was expected, the utilization of the functional units increased since more independent instructions were available. IBM patents and disclosures on multithreading include: US Patent 3,728,692, J.W. Fennel, Jr., Instruction selection in a two-program counter instruction unit, filed August 1971, and issued April 1973. US Patent 3,771,138, J.O. Celtruda, et al.,
Re: DOS descendant still lives was Re: slight reprieve on the z.
t...@vse2pdf.com (Tony Thigpen) writes: > The 4300 did not come out of Endicott. It was developed in Germany, in > the same lab that developes DOS/VSE. As an undergraduate I do lots of work on cp67 (including to run in 256kbyte machine). The morph of cp67 to vm370, did a lot of simplification of cp67 at the same time bloating the kernel size so performance was seriously impacting running in 256kbytes. Boeblingon does 115&125 ... and at one point I get dragged into optimizing vm370 to run on 256kbyte 125 customer machines. boeblingon does 135 and 138. Then Boeblingon does 4331. Endicott had con'ed me into doing lots of work on 148, vm/370 ecps. At the same time I was helping Endicott with 148 vm370 ECPS, the 125 group also asked me to do the design and specification for a 5-way 125 SMP machine (which never shipped, it turns out the Endicott 148 people felt threaten that 5-way 125 multiprocessor would impact their market ... which put me in odd position since I was doing both). Later was dragged into doing a lot of work with regard to 4341. Across the street, the disk engineering group in bldg. 14 and the disk product test group in bldg. 15 dragged me into playing disk engineer ... some past posts http://www.garlic.com/~lynn/subtopic.html#disk the product test group in bldg 15 would typically get the 3rd or 4th engineering model for doing disk i/o testing. they got the 3rd enginneering model of 3033 and very early E4 (4341) engineering machine for testing. Because I was doing so much stuff for them, I would get lots of time on bldg. 15 system for other stuff I might want to do. The performance test marketing group in Endicott con'ed me into doing customer benchmarking on the bldg. 15 enginneering E4/4341 ... since I had bettter access to the machine ... than they had to early engineering machines in Endicott. some old email http://www.garlic.com/~lynn/lhwemail.html#43xx email includes references that when the E4/4341 originally arrived in bldg. 15 ... it had the proceessor cycle slowed down (allowed machine to work as they refined the engineering) so that the benchmarks were not as good as they could be. Later as they refined the machine, they were able to crank down the processor cycle. One of the benchmarks was for LLNL that was looking at getting 70 4341s for a compute farm (sort of the precursor to modern cluster, GRID and supercomputing). It showed 4341 was faster than 158&3031 and 4341 cluster was faster, cheaper, much less floor space and environmentals than 3033. The cluster 4341 threat to 3033 was so big that at one point, the head of POK got corporate to cut the allocation of critical 4341 manufacturing component in half. other trivia: circa 1980, there was plan to move the large variety of internal microprocessors to 801/RISC, including low (vertical microcode) 370s, what was to be as/400, lots of controllers, etc. For various reasons that effort floundered and they continued with various CISC microprocessors. I helped somebody in endicott with white paper showing that VLSI was moving to the point, that large part of 370 could be implemented in silicon ... as opposed to having to be emulated ... which would be much faster & price/performance than pure emulation in 801/RISC (other side effect was that some number of 801/RISC engineers leave and go to work on RISC projects at other vendors). Boeblingon does 4361 (4331 follow-on) and Endicott does 4381 (4341 follow-on) in CISC. IBM was expecting that 4361/4381 would continue the enormous 4331/4341 sales explosion, but by that time the mid-range market was starting to move to workstations and large PCs. Previous posts mentions in the wake of the FS failure, there was mad rush to get stuff back into 370 product pipeline, this included 3033 (168 logic remap to faster chips) and 3081/xa kicked off at the same time http://www.jfsowa.com/computer/memo125.htm Turns out during the 3033 engineering period, I was also involved in a 16-way 370 SMP effort and we con'ed some of the 3033 processor engineers to work on it in their spare time. At first everybody thot it was really great effort, and then somebody tells the head of POK that it could be decades before MVS had effective 16-way support ... and he then invites some of us to never visit POK again ... and tells the 3033 processor engineers to stop being distracted by other activities. With the 3033 out the door, the 3033 processor engineers start work on trout1.5 (aka 3090, in parallel with ongoing 3081/xa) circa 1980. Part of the 3090 effort was to use 4331 as service processor running a highly modified version of vm370 release 6 ... and I periodically get dragged into that effort. The 3090 service processor eventually gets upgraded to pair of 4361s running highly modified version of vm370 release 6 ... a couple (later) old email references http://www.garlic.com/~lynn/2010e.html#email861031 http://www.garlic.com/~lynn/2010e.html#email861223 Early days of REXX (well before ships to
Re: DOS descendant still lives was Re: slight reprieve on the z.
other trivia in the wake of FS and mad rush ... 303x was kicked off ... as mentioned 3033 was 168 logic remapped to 20% faster chips ... that happened to have ten times more circuits per chip. Using original 168 logic, 3033 would have been only 20% faster than 168 (aka 3.6mips). However, some specific logic rework to use the larger circuits per chip got it up to 50% faster than 168 (4.5mips). 158 manufacturing had been enormously automated ... somewhat like what they quote for incremental cost of an automobile rolling off the line. The 158 integrated channel microcode was used for the 303x channel director (158 engine w/o 370 microcode and with the integrated channel microcode). 3031 was two 158 engines, one with just the 370 microcode and a 2nd (channel director) with just just the integrated channel microcode. A 3032 is 168-3 reworked to use 303x channel director for external channels. some benchmark numbers for LLNL ... looking at getting 70 4341s for computer farm (precursor to modern GRID, cloud, and supercomputers) 158 3031 4341 Rain 45.64 secs 37.03 secs 36.21 secs Rain4 43.90 secs 36.61 secs 36.13 secs also times approx; 145168-3 91 145 secs. 9.1 secs 6.77 secs also had run in 35.77 on CDC6600. 158 370 was slower than 3031 because the (single) 158 engine was being shared between the 370 microcode and the integrated channel microcode (which ran even when channels were idle). Part of the original morph from cp67 (and 360/67) to VM370 (multiple 370 models) was vm370 had table of supported 370 models ... with various model characteristics. As part of my moving from cp67 to vm370 ... old email reference: http://www.garlic.com/~lynn/2006v.html#email731212 http://www.garlic.com/~lynn/2006w.html#email750102 http://www.garlic.com/~lynn/2006w.html#email750430 ... I replaced the static table of supported 370 models with dynamic code to determaine the characteristics ... it made it much simpler to deploy a csc/vm production system to different machines (like engineering models) not included in the shipped static table of supported machines. some posts discussion work at the scientific center (why it was called csc/vm). some past scientific center posts http://www.garlic.com/~lynn/subtopic.html#545tech Later I transfer to San Jose research ... on the san jose plant site (accross the street from bldgs. 14&15) and csc/vm morphs into sjr/vm. Old 4341 email about engineering model processor cycle time includes reference to checking the DSPSL value ... which is one of my dynamically determined values ... old reference http://www.garlic.com/~lynn/2006y.html#email790220 from my dynamic adaptive resource manager ... which was guinea pig for starting to charge for system/kernel software (customers referred to as "fair share" since the default resource management policy was "fair share") ... some past posts http://www.garlic.com/~lynn/subtopic.html#fairshare then somebody leaks the benchmark numbers to the press ... and they initially try to blame me ... reference http://www.garlic.com/~lynn/2006y.html#email790226 -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: DOS descendant still lives was Re: slight reprieve on the z.
jcew...@acm.org (Joel C. Ewing) writes: > No (about the "free", not about the "dead for decades"), DOS/VS was the > last really free base (last version Release 34?). Perhaps technically > DOS/VSE was "free", as there didn't appear to be a monthly licensing > charge for DOS/VSE itself (Computerworld, April 30, 1979, p4), but in > the practical sense a production DOS/VSE system was definitely not free > as there were monthly support charges for DOS/VSE and separate monthly > licensing plus support charges for must-have VSE add-on components like > VSE/Power and others. DOS/VSE came out with the IBM 4331 & 4341 > processors in 1979 and supported running in both S/370 mode or the > ECPS:VSE mode supported by the 4300 processor family. various legal actions resulted in 23June1969 unbundling announcement where (application) software & other stuff started to be charged for (however they made the case the kernel software should still be free). some past posts http://www.garlic.com/~lynn/submain.html#unbundle during future system effort 370 efforts were being killed off (lack of 370 products there era is credited with clone processors market foothold). after future system failed ... past posts http://www.garlic.com/~lynn/submain.html#futuresys there was mad rush to get stuff back into 370 product pipeline. POK kick off 3033 (168 logic mapped to faster chips) and 3081/XA in parallel ... reference http://www.jfsowa.com/computer/memo125.htm XA had a lot of extensions tailored for MVS. Endicott did something similar for e-architecture (4331 & 4341) tailored for vs1 In large part a single virtual address space supported as part of the hardware architecture. Rather than having segment & page tables ... there were two new instructions that told the machine what virtual address was at what real address ... and invalidated the virtual address. However there was an enormous explosion in vm/4300 sales (before announce, 4341s were referred to a "E4") ... which required multiple virtual address space ... which met that large number of 4300s ran in 370 mode rather than e-mode. Note that POK had convinced corporate to kill off the vm370 product and move all the development people to POK as part of MVS/XA development (including excuse that MVS/XA would ship on time, if they couldn't get the additional resources). Endicott manage to save the vm370 product mission, but had to reconstitute a development group from scratch. some old 4300 related email http://www.garlic.com/~lynn/lhwemail.html#43xx Note that VS1 and VM/370 "ECPS" was different than e-machine architecture. It originated with the 138/148 (virgil/tully) ... where selected high use kernel/system instructions paths were implemented in microcode. The low & mid-range machine were vertical microcode machines with an avg of 10 native instructions per 370 instruction (somewhat analogous to mainframe emulators that run on Intel platforms). Kernel instruction paths tended to get 10:1 performance improvement when moved to microcode. I did the initial study and effort for the VM/370 ECPS ... old post with results for selecting pathlengths to be moved to microcode (I was told, that I needed to select the 6kbytes of highest executed kernel, which turned out to account of 80% of vm/370 kernel execution) http://www.garlic.com/~lynn/94.html#21 370 ECPS VM microcode assist trivia ... the methodology for selecting the VS1 paths wasn't nearly so rigorous. other trivia ... major motivation for Future System product was as countermeasure to clone controllers ... but the (failed) Future System effort contributed significantly to the rise of the clone processors. The threat of clone processors resulted in decision to transition to charging for system/kernel software. I continued to work on 370 stuff all during the FS period ... even periodically ridiculing FS stuff ... which wasn't exactly career enhancing. Also one of my hobbies was developing advanced enhanced operating systems for internal datacenters ... some old email http://www.garlic.com/~lynn/2006v.html#email731212 http://www.garlic.com/~lynn/2006w.html#email750102 http://www.garlic.com/~lynn/2006w.html#email750430 In any case, the mad rush to get stuff back into 370 product pipeline contributed to decision to pick up various of my stuff and ship it in products for customers. One part of that stuff (dynamic adaptive resource manager) was selected to be guinea pig for starting to charge for system/kernel software ... and I had to spend some amount of time with lawyers & business types going over policies for system/kernel software charging. even more trivia: when 3033 looked at doing something similar to ECPS ... it didn't work out as well. 3033 was horizontal microcode machine that had been optimized so it was executing nearly one 370 instruction per machine cycle. Directly dropping system/kernel 370 pathlengths into microcode could even result in running slower than the original 370. -- virtualization experience
Re: OT: Electrician cuts wrong wire and downs 25,000 square foot data centre
p...@petelancashire.com (Pete Lancashire) writes: > Showing my age > > I worked for Burroughs as an engineering technician. > > A customer with 360/65 instantaneous loss of power. I was there only for a > couple hours to drop off some equipment. Later heard they lost a couple > disk packs. separate from power failures can precipitate disk drive failure. IBM CKD dasd had power loss failure mode ... where there wasn't enough power to maintain memory contents ... but there was enough power left for the controller to complete a write operation ... problem was that the channel had stopped transferring data ... so the controller continued writting all zeros. The result was that after recovery, a subsequent read would show no errors ... for the record that had write operation ("correctly") complete with all zeros (this was especially troublesome when things like VTOC record was being written) FBA introduced that a physical record would not be written unless all data was available to correctly complete a write. This philosophy continued for RAID (write "failure" either completes correctly or at least results in error indication for subsequent read). During the 80s, there was lots of work trying to figure out how to retrofit such a fix to CKD dasd ... or at least provide a way for system to recognize an incorrect trailing zeros write. -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: OT: Electrician cuts wrong wire and downs 25,000 square foot data centre
tony.j.new...@btinternet.com writes: > This happend to us, 3380 continued to write x'00' over VM byte > allocation map on cyl 0. Original CMS filesystem from the mid-60s almost had a fix for this ... updated filesysem control information was written to new locations ... and then the MFD was rewritten pointing to the new version of control information rather than old version. It worked for all writes except for the MFD. The new CMS "EDF" filesystem in the 2nd half of the 70s went to a pair of MFDs. There was the current MFD and a write would always be to the alternate MFD ... if completely corectly, the alternate becomes the (new) current and the (old) current becomes the alternate. A version number goes at the end of (EDF) MFD, on startup/recovery, both MFD is read and the most recent one is used ... a power failure, trailing zeros write would always result in that record appearing as older than the other MFD. In early 80s, I did a CP kernel filesystem ... including spool file system that addressed the problem for CP also. My motivation was that I had HSDT project and I needed VM370 spool file system that ran much faster for RSCS driving multiple T1 (& faster) links (standard spool file system typically got 5-32kbytes/sec ... I needed 3mbytes or better throughput. some past posts http://www.garlic.com/~lynn/subnetwork.html#hsdt I did the implementation in vs/pascal running in virtual address space but still managed to significant improvement over the standard implementation done as part of vm370 kernel in assembler. At one point, I thought I finally had a path to getting it picked up through the corporate network backbone which nodes were moving to multiple 56kbit links. However, this was about the time the communication group was putting intense pressure on the corporate nework to move to SNA ... technical people started being excluded from the backbone meetings ... so they could focus on the pressure being applied to move to SNA. However, by that time I also was doing the throughput enhancements to mainframe TCP/IP product (also implemented in vs/pascal). At the time, the standard product got about 44kbytes/sec using nearly full 3090 processor. In some tuning tests of the "fixes" at Cray Research between Cray and 4341 ... got channel speed throughput using only modest amount of the 4341 processor (possibly 500 times improvement in bytes moved per instruction executed). some past posts http://www.garlic.com/~lynn/subnetwork.html#1044 old email reference to communication group forcing internal network to move to SNA: http://www.garlic.com/~lynn/2006w.html#email870302 and http://www.garlic.com/~lynn/2011.html#email870306 other trivia ... I had also done a paged-mapped CMS filesystem originallyh for CP67 ... and then later moved to VM370. In the early 80s (on 3380 drives), side-by-side comparison of moderate i/o intensive workload ... it would get three times the througput of standard CMS filesystem. some past posts http://www.garlic.com/~lynn/submain.html#mmap old post with some (mmap) benchmark measurements from the 1st half of the 80s ... included in '86 SEAS presentation http://www.garlic.com/~lynn/2006.html#25 DCSS as SWAP disk for z/Linux and repeated in hillgang meeting http://www.garlic.com/~lynn/2011c.html#88 Hillgang -- VM Performance -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: IBM Automatic (COBOL) Binary Optimizer Now Availabile
hyperthreading trivia ... early 70s, I got sucked into helping with hyperthreading effort for 370/195 (that never shipped). 370/195 could run at 10MIPS, but most codes ran at 5MIPs. 195 had our-of-order execution, but didn't have branch prediction or speculative execution ... so conditiional branches stalled the pipeline (had to do careful programming to get 10MIPS, but abundance of conditional branches in most codes would keet machine to 5MIPS). Two i-streams running at 5MIPS would be able to keep machine running aggregate 10MIPS. Idea was to have dual i-stream and registers ... but same single pipeline and execution units ... with instructions flagged as to i-stream. this discussion of end of ACS-360 includes references to hardware muiltithreading patents (and red/blue bit tagging). http://people.cs.clemson.edu/~mark/acs_end.html other acs-360 reference/trivia ... Amdahl says that executives were afraid that it would advance state-of-the-art too fast and company would loose control of the market ... so acs-360 was shutdown. There is also description of acs-360 features that show up over 20yrs later in ES-9000. -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: IBM Automatic (COBOL) Binary Optimizer Now Availabile
charl...@mcn.org (Charles Mills) writes: > Now, in a sense, mainframes ARE getting faster. More cache. Higher > real memory limits and for Z, dramatically lowered memory prices. That > processor multi-threading thing. But especially, new instructions that > are inherently faster than the old way of doing things. Load and store > on condition are the i-cache's dream instructions! Lots and lots of > new "faster way to do things" instructions on the z12 and z13. cache miss access to memory ... when measured in number of processor cycles ... is compariable to 60s disk access time when measured in number of 60s processor cycles. non-mainframe processors have been doing memory latency compensation for decades, out-of-order execution, branch prediction, speculative execution, hyperthreading, etc (aka waiting for memory access increasing being treated like multiprogramming in the 60s while waiting for disk i/o). Also, industry standard, non-risc processors some time ago introduced risc micro-ops ... where standard instructions were translated into risc microops for execution scheduling. mainframe implementations are more & more reusing industry standard implementations, fixed-block disks, fibre-channel standard, CMOS, etc. Half the per-processor performance improvement from z10->z196 playing catchup, is claimed to be introduction of some of these industry standard memory access compensation technologies with further additions in z12 (its not clear about z13 ... some numbers about total system throughput compared to z12, is less than the increase in number of processors ... possibly implying that per processor throughput didn't increase or even declined). -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Were you at SHARE in Seattle? Watch your credit card statements!
martin_pac...@uk.ibm.com (Martin Packer) writes: > Ah Chip & PIN at last. there was a large pilot deployment in the US around the turn of the century ... however it was in the "YES CARD" period ... the issue was it was possible to use the same skimming exploits to collect information for counterfeit magstripe card ... for making counterfeit chipcards. Gov. LEOs did a description of "YES CARD" cases at an ATM Integrity TaskForce meeting ... prompting somebody in the audience to exclaim that they managed to spend billions of dollars to prove that chipcards are less secure than magstripe. In the wake of that, all evidence of the pilot evaporated w/o a trace and speculation was that it would be a long time while things were tried in the US again (waiting for more glitches to be worked out in other jurisdictions). The problem was 1) it was as easy to make counterfeit chipcards as magstipe and 2) they had moved business rules out into the chip. A chipcard terminal would ask the chip 1) was the correct PIN entered, 2) should the transaction be done offline, 3) is the transaction within the credit limit. A counterfeit "YES CARD" would answer "YES" to all three, so didn't need to know the correct PIN and didn't need to do online check with backend (and all transaction are approved). Traditional countermeasure for counterfeit magstripe card is to deactivate the account at the backend ... but that doesn't work with "YES CARD" I had warned the people doing the pilot about the problems, but they went ahead and did it anyway (they were myopically focused on lost/stolen cards and ignored the counterfeit "YES CARD" scenarios). Reference to "YES CARD" presentation at the bottom of this CARTES2002 trip report (gone 404, but lives on at the wayback machine) http://web.archive.org/web/20030417083810/http://www.smartcard.co.uk/resources/articles/cartes2002.html disclaimer: in the mid/late 90s, I was asked to do a protocol that had no such vulnerabilities and was significantly more secure ... then the transit industry also requested that it could also run contactless within the power constraints of transit turnstyle (w/o any reduction in security) ... have you seen how long these transactions take? ... even when they are getting full contact power. -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Were you at SHARE in Seattle? Watch your credit card statements!
slight mainframe related trivia. chip had a booth at the '99 world-wide retail banking conference ... along with press release ... in this old post http://www.garlic.com/~lynn/ansiepay.htm#x959bai X9.59/AADS announcement at BAI leading up to the conference ... we spent a lot of time with one of the other companies including regular meetings with their CEO ... who in prior life had been president of DSD (pok mainframe). a lot of the work had started in the x9a10 financial standard working grop which had been given the requirement to preserve the integrity of the integrity of the financial infrastructure for *ALL* retail payments ... and as a result it was required to not only work for point-of-sale ... but *ALL* payments (including internet). The downside was that eliminating much of the fraud it commoditized payments and reduced barriers to entry. '99 was also the year that GLBA passed (now better known for repeal of Glass-Steagall), rhetoric on the floor of congress was the (original) primary purpose of GLBA was to prevent new entries into banking (especially prevent competition from entities with much more efficient technologies). -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Compiler
000433f07816-dmarc-requ...@listserv.ua.edu (Paul Gilmartin) writes: > Well, yes. Something about core competency. Spend programming > resource on an optimizing compiler which can produce object code > faster, better, cheaper than redundant effort by human programmers. > And the next generation ISA can be exploited merely by recompiling, > not recoding. modern compilers will have detailed knowledge of ISA and lots of programming tricks/optimizations/techniques done by the very best assembler programmers (compiler stat-of-the-art is typically considered having reached this point for most things at least by the late 80s). One of the issues is C language has some ill defined & ambiguous features that inhibits better optimization (that is possible in some better defined languages). minor reference (not only optimization issues but also bugs) http://www.ghs.com/products/misrac.html This flexibility comes at a cost however. Ambiguities in the C language, along with certain syntaxes, consistently trip up even the best programmers and result in bugs. For software developers, this means a large amount of unexpected time spent finding bugs. For managers, this often means the single largest risk to their project. ... snip ... The original mainframe TCP/IP product was done in pascal/vs ... and had none of the programming bugs that have been epidemic in C language implementations. -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Gene Amhdahl Dies at 92
stars...@mindspring.com (Lizette Koehler) writes: > Gene Amdahl, who helped IBM usher in general-purpose computers in the 1960s > and > challenged the company's dominance a decade later with his eponymous machines, > has died. He was 92. > He died on Nov. 10 at Vi at Palo Alto, a continuing care retirement community > in > Palo Alto, California, his wife Marian Amdahl said in a telephone interview. > The > cause was pneumonia, and he had Alzheimer's disease for about five years. end of ACS, Amdahl shutdown ACS-360 http://people.cs.clemson.edu/~mark/acs_end.html ACS was shutdown after ibm management decided it would advance state of the art too fast and they could loose control of the market. Talks about ACS features that finally show up in ES/9000 more than two decades later. Also references multithreading patents. I had gotten sucked into a project that was looking at multithreading 370/195 ... which never shipped. Early 70s, there was FS project that was completely different than 360&370 and was going to completely replace 360/370 ... and internal politics was shutting down 370 efforts. Lack of 370 products during this period is credited with giving clone processors market foothold. http://people.cs.clemson.edu/~mark/fs.html -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Self-service PC
charl...@mcn.org (Charles Mills) writes: > Agreed. I did an HR systems evaluation a few years back (why is a > coder evaluating HR systems? Don't ask.) and all were big on > "self-service," by which they meant if an employee, for example, > wanted to know how many vacation days s/he had in the bank, s/he did > not have to call HR, s/he just signed onto the HR system with a Web > browser (and with "role-based authority" much lower than an HR person) > and looked. 20yrs ago it was webifying callcenter menu screens ... had to have computerized-based authentication front-end and restricting access to information just for the authenticated entity. it has been 20yrs of reducing callcenter use (not having real person at the other end). slight topic drift ... 20yrs ago, consumer dailup online banking operations were making presentations at financial conferences on motivation for moving to the internet; primarily development costs for proprietary modem drivers (at the time >60 drivers were typical) and dialup infrastructure, enormous support costs associated with serial-port modems, etc ... all gets offloaded to ISP. Note at the same time, the commercial dialup online banking operations were saying that they would *NEVER* move to the internet because of a long list of exploits and vulnerabilities (many that presist to this day) ... as an aside, the commercial dialup online banking operations have subsequently moved to the internet anyway. self-service PCs in the past were typically associated with "kiosk", library, etc, public PCs that anybody can walk up to (like store machines looking for stock &/or price check) ... as opposed to the webifying callcenter operations. I had some number of meetings with the NIST rbac people in the 90s http://csrc.nist.gov/groups/SNS/rbac/ at the time, it was much more oriented towards simplifying security office handing out fine-grain access ... and codifying multi-party operations as countermeasure to insider threats (no single person had sufficient authority to complete any high-value operation). -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: More "ageing mainframe" (bad) press.
vbc...@gmail.com (Vince Coen) writes: > I think the stats on migration failures show that many fail regardless > of the target migration mainly is that they over estimate project > time, and quality of the target systems being used in place of m/f. > > Taking a straight view the mainframe is slow compared to running on > servers on a instruction throughput basis. > > What they miss however is the data through put specs compared to > mainframes where the m/f still wins hands down. > > I have tried (just for my self) to build a 8 core PC with separate > Sata controllers for each 15000 rpm drive to match up with m/f > performance but apart from the high costs of each controller there is > still the speed or lack of it of going from the controllers to the > application because of bottle necks in the data bus. > > I have not seen any PC/server design mobo that gets around this > problem and until they do - the mainframe is still "the man" for data > processing in bulk. Lots of migration failures are trying to make any change at all. A simple scenario is the financial industry spent billions of dollars in the 90s to move from "aging" overnight (mainframe) batch settlement to straight-through processing using large numbers of parallel "killer micros". A major source of failure was wide-spread use of industry parallelization libraries (that had 100 times the overhead of cobal batch). I pointed it out at the time, but was completely ignored ... the toy demos looked so neat. It wasn't until they tried to deploy that they ran into the scaleup problems (the 100 times parallelization overhead total swamped the antificapated throughput increases using large number of "killer micros" for straight-through processing). In the meantime there has been enormous amount of work by the industry (including IBM) on RDBMS parallizing efficiencies. A RDBMS-based straight-through processing implementation done more recently easily demonstation all of the original objectives from the 90s ... but the financial industry claimed that it would be at least be another decade before they were ready to try again (lots of executives still bore the scars from the 90s failures and had become risk adverse). In 2009, non-mainframe IBM was touting some of these RDBMS parallelization scaleup efficienices. I somewhat ridiculed them ... "From The Annals of Release No Software Before Its Time" ... since I had been working on it 20yrs earlier (and got shutdown, being told I was not allowed to work on anything with more than four processors). Also, in 1980 I got sucked into to do channel extender for STL that was moving 300 people from the IMS group to off-site bldg. The channel extender work did lots of optimization to eliminate the enormous channel protocol chatter latency over the extended link ... resulting in no appearant difference between local and remote operation. The vendor then tried to get IBM approval for release of my support ... but there was group in POK working on some serial stuff (and were afraid if it was in the market, it would make releasing their stuff more difficult) and managed to get approval blocked. Their stuff is final released a decade later, when it is already obsolete (as ESCON with ES/9000). some past posts http://www.garlic.com/~lynn/submisc.html#channel.extender In 1988, I was asked to help LLNL standardize some serial stuff they have, which quickly morphs into fibre-channel standard (including lots of stuff that I had done from 1980). Later some of the POK engineers define a heavy weight protocol for fibre-channel that drastically reduces the native throughput which is eventually released as FICON. some past posts http://www.garlic.com/~lynn/submisc.html#ficon The latest published numbers I have from IBM is peak I/O benchmark for z196 that used 104 FICON (running over 104 fibre-channel) to get 2M IOPS. At the same time there was a fibre-channel announced for e5-2600 blade that claimed over million IOPS (two such fibre-channel has greater native throughput than 104 FICON running over 104 fibre-channel). In addition, there hasn't been any real CKD manufactured for decades, CKD is simulated on industry standard fixed-block disks. It is possible to have high-performance server blades running native fibre-channel with native fixed-block disks that eliminates the enormous FICON and CKD simulation inefficiencies. Related z196 I/O throughput number is all 14 SAPs running at 100% busy peaks at 2.2M SSCH/sec ... however, they recommend that SAPs are limited to 75% or 1.5M SSCH/sec. I have yet to see equivalent numbers published for EC12 or z13. EC12 press has been that z196 @ 50BIPS processing to EC12 @ 75BIPS processing (50% more processing) only claims 30% more I/O throughput. z13 quote has been 30% more processing than EC12 (with 40% more processors than EC12). Note that while fibre-channel wasn't originally designed for mainframe ... but for non-mainframe server configurations (that tend to run a few thousand),
Re: Setting the writers right
charl...@mcn.org (Charles Mills) writes: > "... the OPM is facing a huge problem with modernizing its security measures > and tactics because of one acronym: COBOL. The programming language that > rose to prominence in the 1960s is rampant throughout the OPM and with the > advanced persistent threats federal agencies are experiencing, it's a > significant vulnerability." > > -- http://fedtechmagazine.com/OPMhack somewhat obfuscation and misdirection away from the massive uptick in outsourcing that occured last decade (even some slightly related to IBM). A big part of the uptick in outsourcing was enormous lobbying done by major private-equity firms on behalf of their acquisitions. "OPM Contractor's Parent Firm Has a Troubled History" https://firstlook.org/theintercept/2015/06/24/opm-contractor-veritas/ when CEO leaves IBM, he goes on to head up major private equity firm http://www.wsj.com/articles/SB1037893592918171788 that does LBO of company that employes Snowden. http://www.investingdaily.com/17693/spies-like-us/ Private contractors like Booz Allen now reportedly garner 70 percent of the annual $80 billion intelligence budget and supply more than half of the available manpower. They're not going away any time soon unless the CIA and NSA want to start over and with some off-the-shelf laptops, networked by the Geek Squad from Best Buy. Security clearances used to be a government function too, but are now a profit center for various private-equity subsidiaries. ... snip ... Private equity tends to do everything possible to loot the companies they acquire (the industry had gotten such a bad name during the S crisis, that they change their name to "private-equity"). In the case of the "subsidiaries" doing outsourced security clearance checks, they were filling out the paperwork, but not actually doing the checks they were being paid for. AMEX had been in competition with KKR for the LBO of RJR and KKR wins. KKR runs into problems with RJR and hires away the president of AMEX to turn it around. https://en.wikipedia.org/wiki/Barbarians_at_the_Gate:_The_Fall_of_RJR_Nabisco IBM has gone into the red and was being re-orged into the 13 "baby blues" in preparation of breakup; the board then hires away the former president of AMEX to resurrect the company and reverse the breakup ... using some of the same measures used at RJR http://www.ibmemployee.com/RetirementHeist.shtml there has been long standing revolving door between gov and beltway bandits and/or wallstreet ... example is recent CIA director resigned in disgrace including slap on the wrist for leaking classified documents ... joins KKR. http://dealbook.nytimes.com/2013/05/30/k-k-r-hires-petraeus/ -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: HP being sued, not by IBM.....yet!
john.archie.mck...@gmail.com (John McKown) writes: > They are probably referring to a z, but doing it in such a way as to > totally disparage it. The fact that the z13 is the fastest microprocessor > currently existed just doesn't penetrate their mind because the original > S/360 was designed in the 1960s. z900, 16 processors, 2.5BIPS (156MIPS/proc), Dec2000 z990, 32 processors, 9BIPS, (281MIPS/proc), 2003 z9, 54 processors, 18BIPS (333MIPS/proc), July2005 z10, 64 processors, 30BIPS (469MIPS/proc), Feb2008 z196, 80 processors, 50BIPS (625MIPS/proc), Jul2010 EC12, 101 processors, 75BIPS (743MIPS/proc), Aug2012 I haven't seen the BIPS numbers for z13 yet, just reference that Z13 has about 30% more throughput than EC12 (with 40% more processors) ... which would be about 100BIPS & about 710MIPS/proc. the claims that half the per processor improvement was introduction of features like out-of-order execution, branch prediction, etc. that have been in other chips for decades. e5-2600v1 blade (about concurrent with z196) 400-500+ BIPS (depending on model); around which IBM had base list price of $1815 or about $3.50/BIPS. However, the large cloud megadatacenters claim that they had been doing their own system assemblies for decades (carefully choosing components for total lifetime costs) or around $1/BIPS. The commoditizing of these systems by the large cloud megadatacenters possibly accounts for IBM unloading that product line (the chip manufactures were saying that they were shipping more processor chips to the large cloud megadatacenters than they were shipping to the brand name system vendors). By comparison z196 works out to $560,000/BIPS (w/o software) and EC12 works out to $440,000/BIPS (aka large cloud megadatacenters at $1/BIPS). A e5-2600v3 blade is rated at 2.5 times a e5-2600v1 blade and a current e5-2600v4 blade is rated at 3.5 times a e5-2600v1 blade ... or over 1.5TIPS. A high density rack of e5-2600v4 blades may have more processing power than the aggregate of all mainframes in the world today. -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: HP being sued, not by IBM.....yet!
0047540adefe-dmarc-requ...@listserv.ua.edu (Bill Johnson) writes: > http://www.theregister.co.uk/2015/09/22/michigan_sues_hp_for_upgrade_failure/ > Michigan failure. remember HP had bought EDS: http://www8.hp.com/us/en/hp-news/press-release.html?id=169924 originally founded by former IBM salesman who then created a mainframe services empire. Along the way, EDS was bought by GM to try and improve their IT position and then later spun off (before being acquired by HP) http://www.dallasnews.com/business/headlines/20121209-eds-sees-success-after-purchase-by-gm--but-at-a-cost.ece trivia: in 1990, GM had the C4 taskforce to look at completely remaking themself ... since they were planning on heavily leveraging IT technology, they invited representatives from technology companies to participate. They could clearly articulate what the foreign competition was doing right and the changes GM would have to make, but as the bailout this century shows they weren't able to make the necessary changes (one of the shortcomings of their operation they described was also true of the mainframe operation at the time ... so I would chide the mainframe representatives how could they expect to contribute). more trivia: congress had passed foreign import quotas to significantly reduce competition and greatly increase domestic profits supposedly that they would use the money to completely remake themselves. However in the early 80s there was a call for 100% unearned profit tax on the US industry since they were just pocketing the profit and continued business as usual (through the 1990 C4 taskforce up through the recent bailouts and possibly beyond, "remake" their business is 35 or so years old, and it hasn't been the case that they didn't know what they needed to do). In the mid-80s we had opportunity to talk to GM/EDS periodically. Then we were doing a presentation in Raleigh on "real" networking and happen to mention that the GM/EDS people mentioned that they were moving the company off SNA to x.25. The communication group executives immediately left the room ... and then came back and said they didn't care, GM/EDS had already spent their IBM communication budget for the year ... so it would be somebody else's problem in the future other trivia: when Perot left GM/EDS and founded Perot Systems, he brought in the former head of POK mainframe to be CEO. past posts mentioning C4-taskforce http://www.garlic.com/~lynn/submisc.html#auto.c4.taskforce -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Term "Open Systems" (as Sometimes Currently Used) is Dead -- Who's with Me?
t...@harminc.net (Tony Harminc) writes: > In my experience, though, Windows was not generally included in what > people meant by "open systems"; they meant UNIX, and if they failed to > include z/OS (or OS/390) UNIX, it's because they were unaware of its > existence. If they wanted to include Windows in a term meaning "not > mainframes", they'd say "distributed systems". I hear very few people > these days use the term "open systems" at all. re: re: http://www.garlic.com/~lynn/2015g.html#77 Term "Open Systems" (as Sometimes Currently Used) is Dead -- Who's with Me? http://www.garlic.com/~lynn/2015g.html#79 Term "Open Systems" (as Sometimes Currently Used) is Dead -- Who's with Me? http://www.garlic.com/~lynn/2015g.html#80 Term "Open Systems" (as Sometimes Currently Used) is Dead -- Who's with Me? google archive https://groups.google.com/forum/#!topic/bit.listserv.ibm-main/dvpRJRmFIJA advent of single chip processors met that companies could develop hardware systems at very low cost but there was still enormous cost associated with developing proprietary operating systems. the thing that unleashed was these companies being able to adapt unix for their hardware at small fraction of developing proprietary operating system from scratch. Saw big explosion in companies doing minis, workstations, mini-supers, supers, etc all using commodity processor chips and portable unix. IBM's office products group was going to use 801/RISC ROMP chip to do a displaywriter followon ... when that got canceled they decided to retarget to the Unix workstation market and got the company that had done the AT unix port for IBM/PC PC/IX to do one for ROMP ... released as PC/RT and AIX2. Some past posts http://www.garlic.com/~lynn/subtopic.html#801 Along the way saw universities doing unix work-alikes ... UCB doing BSD, UCLA doing Locus, CMU doing MACH, etc. IBM Palo Alto Science Center was working on doing UCB BSD for 370 when they got retargeted to PC/RT ... which came out as AOS. They had also been working with UCLA Locus ... which was eventually released as AIX/370 & AIX/386 (Locus AIX having little directly to do with AT UNIX for PC/RT). Jobs had left Apple and was doing NeXT and using MACH as base system, when Jobs comes back to Apple, he brings MACH with him to be the basis for applie operating system. AT & SUN then try and make UNIX more "proprietary" ... kicking off the UNIX wars https://en.wikipedia.org/wiki/Unix_wars and https://en.wikipedia.org/wiki/Open_Software_Foundation The organization was first proposed by Armando Stettner of Digital Equipment Corporation at a by-invitation-only meeting hosted by DEC for several UNIX system vendors in January 1988 (called the "Hamilton Group", since the meeting was held at DEC's offices on Palo Alto's Hamilton Avenue).[3] It was intended as an organization for joint development, mostly in response to a perceived threat of "merged UNIX system" efforts by AT Corporation and Sun Microsystems. ... The foundation's original sponsoring members were Apollo Computer, Groupe Bull, Digital Equipment Corporation, Hewlett-Packard, IBM, Nixdorf Computer, and Siemens AG, sometimes called the "Gang of Seven" ... snip ... which also gave big boost to POSIX https://en.wikipedia.org/wiki/POSIX The disk division executive sponsored POSIX/Open implementation for MVS ... as part of work around communication group opposition to client/server and distributed computing ... but it was also motivated by being able to bid MVS for gov. contracts requiring POSIX compliance. i86 processors have became dominate commodity processor chip ... drastically reducing some of the hardware portability issues ... and windows took over as dominant operating system. rise of Linux was partially because the new computing paradigm for GRID and CLOUD computing that can have millions of processors ... and being able to evolve that new computing paradigm needed full, unrestricted source. Big cluster supercomputers evolved into GRID ... and the big cloud megadatacenters were not too far behind (started leveraging some of the same components) ... mostly dependent on freely available Linux source (although as the paradigm matured, some cases of other systems jumping on the bandwagon). The big cloud megadatacenters have enormously expanded on-demand computing ... there are even instances of dynamically spinning up on-demand supercomputer using credit card. Four years ago there was case of ondemand 240TIPS supercomputer created for research. Year later, there was case of dynamically created ondemand supercomputer that was three times larger, for 3hrs of cancer research (would have ranked in the top 50 supercomputers in the world). By comparison, max configured EC12 is 101 processors rated at 75BIPS (743MIPS/proc), z13 claims 30% more throughput than EC12 with 40% more processors (700MIPS/proc?). 240TIPS would be equivalent of over 3000 max configured EC12 systems ... and more recent, more like 10,000 max
Re: Term "Open Systems" (as Sometimes Currently Used) is Dead -- Who's with Me?
j...@well.com (Jack J. Woehr) writes: > How about "if all my disparate operating systems support TCP/IP and > C/C++, it's easier to accomplish the mission"? > > Which is more or less what it has come down to. re: http://www.garlic.com/~lynn/2015g.html#78 Term "Open Systems" (as Sometimes Currently Used) is Dead -- Who's with Me? original mainframe TCP/IP product was done in vs/pascal for VM370 and communication group was sort of pushed into corner to eventually let it be released ... however with some performance issues (max 44kbyte/sec using nearly full 3090 processor). It was ported to MVS by providing simulation for the required VM370 functions. Open systems have had epidemics of exploits and vulnerabilities attributed to C-language buffer length and addressing semantics (the mainframe vs/pascal implementation was not known to have similar problems) ... some past posts on the subject http://www.garlic.com/~lynn/subintegrity.html#buffer I did the modifications to the vm370 version to support RFC1044 ... and in some tuning tests at Cray research got sustained channel throughput between 4341 and Cray using only modest amount of 4341 processor time. (possibly 500 times improvement in bytes moved per instruction executed). some past posts http://www.garlic.com/~lynn/subnetwork.html#1044 much later the communication group hired a subcontractor to do TCP/IP implementation in vtam and after the initial demonstration they told him that everybody *knows* that a *valid* tcp/ip implementation is slower than LU6.2, and they would only be paying for a *valid* implementation. we were also working with NSF and its supercomputer centers on interconnecting the labs ... originally we were suppose to get $20M. Then congress cut the budget and some other things happened and finally NSF releases RFP (several pieces based on what we already had running) ... but internal politics prevents us from bidding. The director of NSF tries to help and writes the corporation a letter (copying the CEO) with support from other agencies ... but that just made the internal politics worse (as did comments that what we already had running was at least five years ahead of all bid submissions). some old email http://www.garlic.com/~lynn/lhwemail.html#nsfnet as regional networks connect into the nodes, it evolves into the NSFNET backbone ... precursor to modern internet. some discussion http://www.technologyreview.com/featuredstory/401444/grid-computing/ some past posts http://www.garlic.com/~lynn/subnetwork.html#nsfnet and http://www.garlic.com/~lynn/subnetwork.html#internet -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Term "Open Systems" (as Sometimes Currently Used) is Dead -- Who's with Me?
j...@well.com (Jack J. Woehr) writes: > Not Found ... but I went through several of the others ... > one could spend the rest of one's careers reading your posts ;) re: http://www.garlic.com/~lynn/2015g.html#77 Term "Open Systems" (as Sometimes Currently Used) is Dead -- Who's with Me? http://www.garlic.com/~lynn/2015g.html#79 Term "Open Systems" (as Sometimes Currently Used) is Dead -- Who's with Me? garlic.com changed/moved their webserver on 16apr2015 and I am still trying to work out how to update files at the new webserver (i've exchange lots of email, still haven't worked it out). in mean time, this thread archived in google groups https://groups.google.com/forum/#!topic/bit.listserv.ibm-main/dvpRJRmFIJA reference here from last decade http://www.ibmsystemsmag.com/mainframe/stoprun/Stop-Run/Making-History/ I was blamed for online computer conferencing on the internal network (larger than arpanet/internet from just about the beginning until mid-80s) in the late 70s & early 80s; folklore is that when the corporate executive committee was told about online computer conferencing (and the internal network), 5of6 wanted to fire me. one reference (from IBMJargon): Tandem Memos - n. Something constructive but hard to control; a fresh of breath air (sic). That's another Tandem Memos. A phrase to worry middle management. It refers to the computer-based conference (widely distributed in 1981) in which many technical personnel expressed dissatisfaction with the tools available to them at that time, and also constructively criticised the way products were are developed. The memos are required reading for anyone with a serious interest in quality products. If you have not seen the memos, try reading the November 1981 Datamation summary. ... snip ... somewhat as a result, a researcher was paid to sit in the back of my office for 9months to study how I communicated; took notes on face-to-face and telephone, went with me to meetings, got logs of my instant messages and copies of all incoming & outgoing email. material was used for research report, some number of papers & books and a Stanford PHD (joint between language and computer AI). some past posts http://www.garlic.com/~lynn/subnetwork.html#cmc (children's "bullying") book about former co-worker at the science center responsible for the internal network http://itunes.apple.com/us/app/cool-to-be-clever-edson-hendricks/id483020515?mt=8 It's Cool to Be Clever: The Story of Edson C. Hendricks, the Genius Who Invented the Design for the Internet http://www.amazon.com/Its-Cool-Be-Clever-Hendricks/dp/1897435630/ and wiki https://en.wikipedia.org/wiki/Edson_Hendricks the internal network technology was also used for the corporate sponsored university network ... also larger than internet for a time (and where ibm-main originated) ... wiki reference http://en.wikipedia.org/wiki/BITNET some past posts http://www.garlic.com/~lynn/subnetwork.html#bitnet some past science center posts http://www.garlic.com/~lynn/subtopic.html#545tech -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Term "Open Systems" (as Sometimes Currently Used) is Dead -- Who's with Me?
imugz...@gmail.com (Itschak Mugzach) writes: > The term 'open' for me is the liberty to choose. To choose the > hardware from many makers and to move easily from one operating system > to another. See how many are moving from unix to Linux so easy. The > mainframe is not dead nor the customers. They, who can choose, vote > for liberty. IBM killed all alternative hardware makers and now they > buy or resell software makers. In Israel most of the mainframe sites > are looking their way out, telling them selves this is because COBOL, > pricing, and other tails. The real truth is that IBM kills the > industry by being a monopoly. The only chance I see for the industry > is IBM allowing alternatives. I had the discussion with the (disk division) executive that sponsored the "open" implementation for MVS. His view would be that it would make it easy for customers to move applications from non-IBM platforms to MVS. My claim was industry motivation for "open" was that it would make it trivial for customers to move applications to whatever platform they wanted to ... eliminating proprietary lockins that had previously been the norm. Frequently move to the latest most cost effective platform, aka hardware platform agnostic, promoting competition and accelerating improvements. It is also seen in things like industry standard benchmarks like TPC (price/transaction, watts/transaction, etc) http://www.tpc.org/ trivia ... former co-worker at ibm san jose research http://www.tpc.org/information/who/gray.asp For IBM, it was sort of in the genre of SAA ... billed as application could be run anywhere ... but really met that application could be moved to the mainframe ... while human interface could be on some other platform. SAA was part of communication group desperately fighting off client/server and distributed computing trying to preserve their dumb (emulated) terminal paradigm and install base. The communication group had stranglehold on datacenters with corporate strategic ownership of everything that crossed the datacenter walls and the disk division was starting to see the effects with drop in disk sales (data fleeing the datacenter to more distributed computing friendly platforms). The disk division had come up with number of solutions to correct the problem ... but the ones that involved actually crossing the datacenter walls were constantly vetoed by communication group. some past posts (including references to claim that communication group was going to be responsible for demise of disk division) http://www.garlic.com/~lynn/subnetwork.html#terminal My wife had written 3-tier architecture into response to large gov. super-secure, campus environment RFI ... and then we were using 3-tier in customer executive presentations and taking all sorts of arrows in the back from the SAA forces. past posts http://www.garlic.com/~lynn/subnetwork.html#3tier The "open" for MVS didn't actually involve anything that crossed the datacenter walls so the communication group couldn't veto it. He was also investing in companies that produced mainframe products that crossed datacenter walls (communication group could veto his developing and selling IBM distributed products that physically crossed datacenter walls, but couldn't veto him from investing in non-IBM companies). -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Mainframes open to internet attacks?
mike.a.sch...@gmail.com (Mike Schwab) writes: How about Multics? Designed from the start to be multi-user and highly secure. some of the CTSS people went to the 5th flr and did Multics. Other of the CTSS people went to the IBM science center on the 4th flr and did cp67/cms, the internal network, online services, etc. Being in the same bldg. separated by one flr, there was some rivalry. One of the early tests was when science center ported apl\360 to cms for cms\apl ... it allowed typical apl\360 16kbyte workspaces to be increased to virtual memory size ... and also added API that allowed access of system services (like file read/write). Opening APL to real-world applications attracted a lot of internal locations to start using the cambridge system remotesly. A group of business planners in Armonk loaded the most valuable corporate asset (customer details) on cambridge system to do business modeling applications in cms\apl. we had some interesting issues since non-employess (cambridge area univ students, instructors, professors) also had online access to the cambridge system. some posts mentioning science center http://www.garlic.com/~lynn/subtopic.html#545tech some multics installations: http://www.multicians.org/site-afdsc.html http://www.multicians.org/mgd.html#DOCKMASTER other old reference to DOCKMASTER org. (gone 404 but lives on at wayback machine): http://web.archive.org/web/20090117083033/http://www.nsa.gov/research/selinux/list-archive/0409/8362.shtml and old reference to afds coming by to talk about 20 vm/4341 systems ... but then that was increased to 220 (posted in multics discussion group) http://www.garlic.com/~lynn/2001m.html#email790404 Recently a european that worked in NATO claimed that they got 6000 vm/4341 systems. Note that Multics was implemented in PLI. Up through the 90s, the major tcp/ip bugs/exploits were because of buffer length related bugs epidemic in c-language implementations (and still continues to be a frequent source of exploits). The original ibm mainframe tcp/ip product was implemented in vs/pascal and had *none* of these epidemic bugs found in c-language implementations. As an aside, for various reasons this implementation had some significant performance issues, getting 44kbytes/sec aggregate using 3090 processor. I did the rfc1044 enhancements and some tuning tests at cray research got sustained channel speed throughput between cray and 4341, using only modest amount of 4341 (possibly 500 times improvement in bytes moved per instruction executed). The (non-rfc1044) version was also made available on MVS by simulating the required VM functions. Much later the communication group contracted for TCP/IP support through VTAM. After the initial demonstration, the communication group told the contractor that everybody *knows* that a *correct* version of TCP/IP runs slower than LU6.2 and they will only be paying for a *correct* version. I also had other rivalry with the 5th flr. One of my hobbies was providing enhanced operating systems to internal locations ... some old email regarding CSC/VM (later it was SJR/VM, after I transferred to san jose research): http://www.garlic.com/~lynn/2006w.html#email750102 http://www.garlic.com/~lynn/2006w.html#email750430 It wasn't fair to compare the total number of Multics systems that had ever existed with the total number of vm370 customer systems or even the total number of internal vm370 systems. However, for a time, I had a few more internal csc/vm systems than the total number of Multics systems. -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Miniskirts and mainframes
JimP solosa...@gmail.com writes: Interesting. The main contractor told us it was due to the teraflops it could do, a YMP-2. I worked for a sub-contractor. IBM Kingston supposedly had the responsibility for doing new supercomputer ... also was providing to Chen's endevor (responsible for both xmp ymp) https://en.wikipedia.org/wiki/Steve_Chen_%28computer_engineer%29 we were doing cluster scaleup as part of HA/CMP ... past posts http://www.garlic.com/~lynn/subtopic.html#hacmp then end of Oct91, the senior executive backing Kingston effort retires and there is audit of all his projects. After that they start scouring the company looking for high performance technology. Jan1992 meeting in ellison's conference room about (commerical) cluster HA/CMP scaleup http://www.garlic.com/~lynn/95.html#13 mainframe DB2 complaining if we were allowed to go ahead, it would be years ahead of them. Then cluster scaleup is transferred, announced as supercomputer (for technical scientific *ONLY*) and we were told that we can't work on anything with more than four processors (which motivates us to leave). Somewhat no brainer operations ... it takes care of the DB2 complaints and also gets them their supercomputer ... old email about working on technical and scientific with national labs and others (up until the transfer) http://www.garlic.com/~lynn/lhwemail.html#medusa 17Feb1992 press, scientific and technical *ONLY* http://www.garlic.com/~lynn/2001n.html#6000clusters1 05May1992 press, total surprise about national lab interest http://www.garlic.com/~lynn/2001n.html#6000clusters2 I would claim that (national lab) scientific technical activity traces back to at least getting roped into doing LLNL benchmark looking at getting 70 4341s for computer farm and RDBMS/commercial dates back to BofA getting 60 4341s for branch office distributed System/R (original relational/sql implementation). Note that IBM's RDBMS were mainframe only and so RS/6000 had to work with other RDBMS for their platform. It turns out that a couple of those vendors had same source base for both VMS cluster as well as open systems. Part of simplifying the HA/CMP cluster scaleup was providing interface that supported VMS cluster semantics. These RDBMS vendors also had some very strong feeling about some of the VMS cluster implementation that could be done much better (which I was able to take advantage of besides my experience doing mainframe tightly-coupled and loosely-coupled implementations). Later in the 90s, Chen is CTO at Sequent and we do some consulting work for him (before IBM buys Sequent and shuts it down). -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Formal definituion of Speed Matching Buffer
shmuel+ibm-m...@patriot.net (Shmuel Metz , Seymour J.) writes: I'm editing the wikipedia article on Count Key Data, and I've run into an editorial dispute. I claim that what is now ECKD was part of the SMB, and the other editor claims that you could run 3380 on a slow channel without using, e.g., Define Extent. Does anybody have a document outlining what IBM included in the term SMB?: Thanks. re: http://www.garlic.com/~lynn/2015f.html#86 Formal definituion of Speed Matching Buffer http://www.garlic.com/~lynn/2015f.html#88 Formal definituion of Speed Matching Buffer http://www.garlic.com/~lynn/2015f.html#89 Formal definituion of Speed Matching Buffer http://www.garlic.com/~lynn/2015g.html#4 3380 was actually FBA? http://www.garlic.com/~lynn/2015g.html#6 3380 was actually FBA? http://www.garlic.com/~lynn/2015g.html#9 3380 was actually FBA? need to have calypso (SMB) installed to attach to channel slower than device ... however it is possible to do channel programs w/o using ECKD CCWs GA26-1661-9 3880 Storage Control Description ... from bitsavers 4-23 Define Extent (also 4-26 Locate Record) Note: This command is valid only when the speed matching buffer for the 3375 or 3380 feature is installed. ... snip ... 4-111 describes how speed matching buffer uses the define extent locate record to calculate when to connect to channel ahead of time on write operation. 5-17 I/O Operation for Speed Matching Buffer The speed matching buffer for the 3375 feature and the speed matching buffer for the 3380 feature (Models AA4, A04, and B04 only) will correctly execute standard command chains when connected to channels slower than the 3375 or the 3380. However, a performance reduction on write operations (described below) will occur. ... snip ... -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: 3380 was actually FBA?
l...@garlic.com (Anne Lynn Wheeler) writes: hardware speed and error correction was going to fixed-sized blocks. You can see this in 3380 track capacity calculations where record sizes have to be rounded up, sort of compromise hack given that MVS wasn't going to support real FBA. The 3380 was smaller fixed-sized blocks ... but not true IBM FBA like 3310 3370. 3375 was the first CKD emulated on top of an IBM FBA (3370) device. 512-byte blocks have prevailed for a couple decades (IBM 3310 3370 and follow-ons ... but also all the other industry standard disks). There is currently inudstry move to 4096-byte fixed blocks for improved error correction and track capacity. https://en.wikipedia.org/wiki/Advanced_Format http://www.seagate.com/tech-insights/advanced-format-4k-sector-hard-drives-master-ti/ re: http://www.garlic.com/~lynn/2015g.html#4 3380 was actually FBA? and https://groups.google.com/forum/#!topic/bit.listserv.ibm-main/3QSdKeko604 IBM journal articles are behind IEEE membership wall ... have found this detailed description at Google Books (3380 error correcting) https://books.google.com/books?id=cG4Zgb8OqwECpg=PA495lpg=PA495dq=ibm+3380+error+correctingsource=blots=lMaYN_d94Fsig=o-R202AspjC1Ox09YNcZDb9Ljgchl=ensa=Xved=0CDgQ6AEwBGoVChMIxpHRg-KmxwIVVluICh1twgJy#v=onepageq=ibm%203380%20error%20correctingf=false which has each subblock consists of 96 data bytes and six first-level check bytes are appended in the form of two interleaved codewords after discussing details of 3380, it moves into RAID Reed-Solomon codes ... trivia, I worked with somebody in bldg14 that was awarded the original RAID patent. -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: 3380 was actually FBA?
jcal...@narsil.org (Jerry Callen) writes: In another thread, l...@garlic.com wrote: ... but then if MVS had FBA support wouldn't have needed to do 3380 as CKD (even tho inherently it was FBA underneath) ... I didn't know that. Was that the first (and/or last?) IBM SLED to be inherently FBA under the hood? Where were the smarts for that implemented, in the control unit, or the drive itself? re: http://www.garlic.com/~lynn/2015f.html#86 Formal definituion of Speed Matching Buffer hardware speed and error correction was going to fixed-sized blocks. You can see this in 3380 track capacity calculations where record sizes have to be rounded up, sort of compromise hack given that MVS wasn't going to support real FBA. The 3380 was smaller fixed-sized blocks ... but not true IBM FBA like 3310 3370. 3375 was the first CKD emulated on top of an IBM FBA (3370) device. 512-byte blocks have prevailed for a couple decades (IBM 3310 3370 and follow-ons ... but also all the other industry standard disks). There is currently inudstry move to 4096-byte fixed blocks for improved error correction and track capacity. https://en.wikipedia.org/wiki/Advanced_Format http://www.seagate.com/tech-insights/advanced-format-4k-sector-hard-drives-master-ti/ eckd originally for speed-matching buffer ... was also trying to retrofit a little of the FBA benefits to CKD architecture (again because MVS wouldn't upgrade to real FBA). part of the issue for 3375 was there wasn't a mid-size CKD disk (just the high-end 3380). Large customers were buying hundreds ( thousands) of vm/4300s for distributed non-datacenter market (sort of leading edge of the coming distributed computing tsunami; for instanceNATO got 6000 vm/4341s) ... and MVS couldn't play in that new market with no mid-range CKD disk. Doing the 3375 CKD at least gave MVS a path ... however MVS support was really oriented around having 10-20 people in large datacenter. The idea of supporting a thousand distributed systems out in departmental areas wasn't very practical. I also got dragged into doing benchmarks for LLNL that was looking at 70 4341s for computer farm (sort of leading edge of the future supercomputing paradigm). 4341 was faster than 1583031 ... and clusters of 4341s were faster than 3033, lower aggregate cost, lower aggregate physical and environmental footprint, also higher aggregate memory and i/o throughput. old email http://www.garlic.com/~lynn/lhwemail.html#4341 past posts http://www.garlic.com/~lynn/submain.html#dasd -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Formal definituion of Speed Matching Buffer
re: http://www.garlic.com/~lynn/2015f.html#86 Formal definituion of Speed Matching Buffer http://www.garlic.com/~lynn/2015f.html#88 Formal definituion of Speed Matching Buffer For those that got post forwarded and can't see the recent URL refs on garlic.com ... On 17Apr2015, garlic.com changed there webserver and still have yet been able to update my personal web pages ... to see the full thread check google group archive. https://groups.google.com/forum/#!topic/bit.listserv.ibm-main/K2Elt-40-VE With regard to MVS MTBF 15min, I happened to mention in an internal (only) report giving technical details of building a bullet proof and never fail I/O subsystem ... which brought down wrath of the MVS organization on my head ... apparently they would have gotten me fired if they couldn't figure out how ... but they found other ways of taking out their displeasure. VM370 official calypso support, put8201, reference to retrofitting to heavily modified internal system. To: wheeler Date: 01/11/82 12:51:34 ref calypso-vm/sp (ext. ckd ) ; The official release of the code was this month. Put tape 8201 lvl. 110. It looks like it might be awhile before we get that far for common and all, right ? Maybe I should go ahead and work on fitting it to 106. What do you think? ... snip ... from vmshare archives ... discussion of 3880 speed matching buffer but doesn't explicitly say anything about define extent and eckd http://vm.marist.edu/~vmshare/browse?fn=3380ft=MEMO old email about FE error injection regression tests for 3380 ... all of the 57 errors were resulting in MVS hanging and requiring reipl and in 2/3rds of the cases, there was no indication of what caused the problem (this is separate from the earlier issue where they attempted to use MVS in the bldg. 14 engineering test lab and found it had 15min MTBF) http://www.garlic.com/~lynn/2007.html#email801015 -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Formal definituion of Speed Matching Buffer
shmuel+ibm-m...@patriot.net (Shmuel Metz , Seymour J.) writes: I'm editing the wikipedia article on Count Key Data, and I've run into an editorial dispute. I claim that what is now ECKD was part of the SMB, and the other editor claims that you could run 3380 on a slow channel without using, e.g., Define Extent. Does anybody have a document outlining what IBM included in the term SMB?: Thanks. Calypso (speed-matching) feature for 3880 controller allowing 3380s to be attached to slower channels. I have old email about significant software hardware debugging getting it to work. http://www.garlic.com/~lynn/2007f.html#email801010 http://www.garlic.com/~lynn/2007e.html#email820907 http://www.garlic.com/~lynn/2007e.html#email820907b We had done a VM370 modification at SJR that did super efficient trace of all disk record accessed (both by vm370 virtual machines) what was installed in several systems in the San Jose/Bay Area. the 10Oct1980 email refers to upgrading to support calypso/eckd CCW. The trace was used for various things like modeling disk i/o cache configurations. We had a proposal to have it incorporated into all systems for use in dynamic load-balancing for placement/location. some of the ECKD intertwinces with my theme (rant) that it would have been enormously simpler less expensive to have added FBA support to MVS. I had been told that even if I provided them with fully integrated and tested MVS FBA support ... that i needed a $26M incremental business case (to cover documentation and education) ... basically $200M-$300M in additional disk sales ... but they claimed that customers were already buying disks as fast as they could be made ... so it would just shift the same amount of sales from CKD to FBA (and therefor it was impossible for me to show incremental/additional disk sales from FBA support). http://www.garlic.com/~lynn/submain.html#dasd this post references that 4341 with small tweak was being used for testing 3mbyte/sec channel (w/o needing speed matching) http://www.garlic.com/~lynn/2010e.html#36 What was old is new again (water chilled) If POK machines (158, 168, 303x) had been as powerful as 4341 ... there wouldn't have needed Calypso (speed matching and enormous resources needed to get it working) ... but then if MVS had FBA support wouldn't have needed to do 3380 as CKD (even tho inherently it was FBA underneath) ... or do a 3375/CKD version of 3370/FBA. other past posts mentioning calypso http://www.garlic.com/~lynn/2004o.html#7 Integer types for 128-bit addressing http://www.garlic.com/~lynn/2007e.html#40 FBA rant http://www.garlic.com/~lynn/2007f.html#0 FBA rant http://www.garlic.com/~lynn/2008q.html#40 TOPS-10 http://www.garlic.com/~lynn/2009k.html#44 Z/VM support for FBA devices was Re: z/OS support of HMC's 3270 emulation? http://www.garlic.com/~lynn/2009p.html#11 Secret Service plans IT reboot http://www.garlic.com/~lynn/2010e.html#36 What was old is new again (water chilled) http://www.garlic.com/~lynn/2010h.html#30 45 years of Mainframe http://www.garlic.com/~lynn/2010n.html#14 Mainframe Slang terms http://www.garlic.com/~lynn/2011e.html#35 junking CKD; was Social Security Confronts IT Obsolescence http://www.garlic.com/~lynn/2012j.html#12 Can anybody give me a clear idea about Cloud Computing in MAINFRAME ? http://www.garlic.com/~lynn/2012o.html#64 Random thoughts: Low power, High performance http://www.garlic.com/~lynn/2014m.html#154 BDW length vs. Physical Length -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Formal definituion of Speed Matching Buffer
000433f07816-dmarc-requ...@listserv.ua.edu (Paul Gilmartin) writes: Circa 1980 my then employer marketed a CCD SSD product which suffered timing incompatibilities, not because of transfer rate, but because of inter-block latencies. It appeared that some VM paging code paths depended on completing while the inter-block gap was passing. re: http://www.garlic.com/~lynn/2015f.html#86 Formal definituion of Speed Matching Buffer an issue would be if you simulated 3330 .. and formated 50byte (or 110byte) dummy records and had a simulated rotational spin rate much faster than 3330 (it wasn't vm code paths, it was speed of chained CCW channel processing). I periodically mention getting to play disk engineer in bldgs 1415 (sometimes they demanded I play disk engineer) ... past posts http://www.garlic.com/~lynn/subtopic.html#disk vm370 formated 3330s for paging so records were aligned on each track ... three 4kbyte pages per track ... and if there were queued requests for records on same cylinder but different tracks would attempt to optimize single channel program to transfer all pages in the minimum number of revolutions. For CKD this would require seek head, search, tic, read/write ... and in order to allow the channel time to execute the CCWs while the disk was spinning, page area formating inserted dummy records between page data records. It turns out that channel specs (worst case 370) required 110byte dummy records given the 3330 rotational spin rate ... to allow channel time to process the CCWs (to do a head switch) ... but track size only had room for 50byte dummy records. I did a test program that was run on 145, 148, 4341, 158, 168, 303x with IBM disk controllers and various non-IBM disk controllers. IBM disk controller could actually process the head switch CCWs with 50-byte dummy record (w/o additional revolution), for most machines except 158 all 303x ... and some number of non-IBM disk controllers could do the switch even with 158 303x (the issue is that for all models of 303x, they used external channel director ... which was actually a 158 engine with the slow integrated channel microcode and w/o the 370 microcode). 3081 channels also had problem doing CCW head-switch within the 50byte dummy record window. past posts on the subject http://www.garlic.com/~lynn/2000d.html#7 4341 was Is a VAX a mainframe? http://www.garlic.com/~lynn/2001j.html#3 YKYGOW... http://www.garlic.com/~lynn/2002b.html#17 index searching http://www.garlic.com/~lynn/2003g.html#22 303x, idals, dat, disk head settle, and other rambling folklore http://www.garlic.com/~lynn/2004d.html#64 System/360 40 years old today http://www.garlic.com/~lynn/2004d.html#65 System/360 40 years old today http://www.garlic.com/~lynn/2004d.html#66 System/360 40 years old today http://www.garlic.com/~lynn/2005p.html#38 storage key question http://www.garlic.com/~lynn/2005s.html#22 MVCIN instruction http://www.garlic.com/~lynn/2006w.html#8 Why these original FORTRAN quirks? http://www.garlic.com/~lynn/2008l.html#83 old 370 info http://www.garlic.com/~lynn/2011.html#65 Speed of Old Hard Disks http://www.garlic.com/~lynn/2013e.html#61 32760? http://www.garlic.com/~lynn/2014k.html#26 1950: Northrop's Digital Differential Analyzer There was a different issue with code paths on 3880 controller. After FS death, there was mad rush to get stuff back into 370 product pipelines (during FS period, internal politics had been killing off 370 efforts) and 303x and 370-xa were kicked off in parallel. 370-xa became none as 811 for the Nov1978 date on most of the architecture documents. When I saw SSCH, I thot that it was mostly to compensate for the enormous interrupt pathlength in MVS. A big problem was that as devices became faster, and load increased, there was significant increasing device idle time while MVS went thru interrupt redrive overhead. Earlier, the disk engineering lab had been testing using prescheduled, stand-alone mainframe dedicated test time ... at one point they had tried to use MVS ... but found it had 15min MTBF (requiring manual re-ipl) in that environment. I volunteered to rewrite I/O supervisor so it was bullet-proof and never fail so that any number of on-demand, concurrent testing could go on (vastly increasing productivity). I also setout to demonstrate the optimal interrupt processing and queued request device redrive. some bean counter had dictated that 3880 use a really slow control processor (compared to 3830) and used dedicated circuits to get 3mbyte/sec transfer rate. The slow control processor showed up in increased channel and controller busy as well as increase elapsed time for channel program processing. To pass product acceptance test (requiring 3880 appear to be within 5% of 3830 performance), they would signal channel program complete (CE+DE) early ... before having actually finished everything. The first time they put 3880 controller into use with 16 3330s (and heavy load on 3033) replacing 3830 ...
Re: Limit number of frames of real storage per job
allan.stal...@kbmg.com (Staller, Allan) writes: There can also be performance advantages from GC. GC moves objects together in storage, making it much more likely that your application data will be in the processor caches. If GC keeps your data in processor cache it will perform much better than if it's scattered across a GB of storage. apl\360 would allocate new storage for every assignment statement, quickly using every available location in workspace ... and then it would collect everything in contiguous storage (garbage collect) and then start all over again.. This wasn't too bad with apl\360 typically 16kbyte (sometimes 32kbyte) workspaces there were swapped as integral unit. the initial port of apl\360 to cp67/cms for cms\apl was something of a problem because it allowed workspaces that were the size of virtual memory ... and strategy would quickly result in page thrashing (repeatedly touching every virtual page regardless of actual programdata size). before release of cms\apl, this all had to be reworked in order to reduce the massive page thrashing. Besides doing virtual machines, cp67/cms, the technology for the internal network (and corporate sponsored univ bitnet ... where ibm-main originated), GML, and lots of other things ... the science center http://www.garlic.com/~lynn/subtopic.html#545tech also did a number of performance analysis tools. One did processing storage use analysis ... which was used for analyzing cms\apl and bunch of other things. It was also used extensively inside ibm by most product groups in their transition to virtual memory operation (would identify hot-spot instruction use as well as hot-spot storage use) ... and eventually released to customers as VS/Repack (which attempt semi-automated program reoganization to improve operation in virtual memory environment). references to internal network http://www.garlic.com/~lynn/subnetwork.html#internalnet references to bitnet http://www.garlic.com/~lynn/subnetwork.html#bitnet references to gml (sgml, html, etc) http://www.garlic.com/~lynn/submain.html#sgml a major factor in the motivation in transition from os/360 MVT to virtual memory OS/VS2 was significant problems with the way MVT managed real storage, GETMAIN, etc ... regions had to typically be four times larger than really needed. The analysis showed that typical 370/165 MVT 1mbyte machine only supported four regions. A virtual memory MVT on 370/165 1mbyte machine could support 16 regions with little or no paging (aka keep all the in-use data in the 370/165 1mbyte processor cache). Old reference to study motivating to move all 370 to virtual memory: http://www.garlic.com/~lynn/2011d.html#73 Multiple Virtual Memory -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Where are Internet Lists for Mainframe
stars...@mindspring.com (Lizette Koehler) writes: For those of you going to share in Orlando, I would like to let you know that at Share Tom Conley will be giving a share presentation on Thursday 3:15p called Effective Use of the Internet for Mainframe Problem Solving This session will show better ways of posting and getting responses from various Lists. As well as providing a list of Mainframe specific Lists. a co-worker at the ibm cambridge science center was responsible for the technology for the internal network (larger than the arpanet/internet from just about the beginning until sometime late 85 or early 86) and later bitnet (corporation sponsored univ. network) starting in the early 80s. https://en.wikipedia.org/wiki/Edson_Hendricks I was blamed for online computer conferencing on the internal network in the late 70s early 80s. folklore is that when the executive was told about online computer conferencing (and the internal network) 5of6 wanted to fire me. this is email from person in paris given job of doing EARN (bitnet equivalent in europe) looking for online applications http://www.garlic.com/~lynn/2001h.html#email840320 this is history of listserv creation in paris in 1986 ... somewhat similar to the earlier internal online VMTOOL computer conferencing tool (developed in the wake of my activities) ... although VMTOOL had both usenet-like mode in addition to mailing list like mode. http://www.lsoft.com/corporate/history-listserv.asp http://www.lsoft.com/products/listserv-history.asp wiki https://en.wikipedia.org/wiki/LISTSERV In the late 80s, the communication group was spreading misinformation as part of convincing corporate to ocnvert the internal network to SNA/VTAM ... when it would have been much more efficient to have converted to TCP/IP (like BITNET did). http://en.wikipedia.org/wiki/BITNET more mailing lists http://www.lsoft.com/lists/list_q.html some of the mailing lists gateway to usenet ... some bidirectional, others only distribution (but not posting). usenet is also gatewayed (usually bi-directional) to google groups note that TYMSHARE had developed CMS-based online computer conferencing system and made it avaialble to IBM SHARE organization starting in Aug1976 ... archive: http://vm.marist.edu/~vmshare ... for mainframe specific try search engine for ibm mainframe discussion group mailing list -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: 3705
g...@ugcs.caltech.edu (glen herrmannsfeldt) writes: OK, I forgot that the Usenet gateway doesn't work anymore. I am wondering what software one needs for a 3705 to connect up ordinary ASCII terminals. For example, what would be needed to use TSO or Wylbur on ASCII terminals? I know this is what was done 35 years ago, but I don't know now who knows how to do it. I do remember that for dial-up lines it would allow for 300 baud or 110 baud, or even for 2741s, depending on the first character you typed. Hardwired lines were fixed speed, and could be higher than 300. (I believe O for 300 baud, and S for 110 baud.) Faster lines might only be at a fixed baud rate. cp67 delivered to the univ. had automatic terminal type identification for 1052 and a couple of 2741 types. 2702/2703 was possible to dynamically change the line-scanner type using the SAD CCW (use one line-scanner type, try a couple operations and if they get errors, switch to a different line-scanner type). the univ. had a number of TTY/ASCII so I had to add TTY support to CP67 ... and tried to do it also using dynamic terminal type identification. I also tried to support single dial-in number for all terminal types ... aka hunt group ... common pool of lines. However, IBM had taken short cut and hard-wired the line-speed oscillator to each line ... so while it was possible to change the line-scanner ... it wasn't possible to change the line-speed (original 1052 2741 had same line speed, but TTY was different). This was motivation for univ. to start a clone-controller project, building channel interface board for Interdata/3 programed to emulate 2702 ... but able to also do dynamic line-speed operation. This was later improved to Interdata/4 for the channel interface and cluster of Interdata/3s dedicated to line-scanner. Four of us get written up as responsible for (some part of) the clone-controller business. Later Perkin-Elmer buys Interdata and the clone-controller continues to be sold under the PE logo (in the late 90s, I ran into PE box in large datacenter handling much of the dial-up point-of-sale terminals on the east coast, 1200 baud ascii). A number of univ. had been sold (virtual memory) 360/67s supposedly for use with TSS/360 ... however TSS/360 had hard time reaching maturity ... so a lot of places ran CP/67. Other places developed their own virtual memory operating systems for 360/67 ... Stanford did Orvyl/Wylbur (Wylbur later ported to MVS) and Michigan did MTS. MTS did clone-controller using PDP8 http://www.eecis.udel.edu/~mills/gallery/gallery7.html some more MTS http://www.eecis.udel.edu/~mills/gallery/gallery8.html -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Why major financial institutions are growing their use of mainframes
marktre...@gmail.com (Mark Regan) writes: I recently learned about a bank in Japan that has been using a mainframe since the 1970's without a single second of downtime. Its architecture allows for full software and hardware upgrades without an outage. i periodically mention that my wife had been con'ed into going to POK to be in charge of loosely-coupled architecture where she did peer-coupled shared data architecture. some past posts http://www.garlic.com/~lynn/subnetwork.html#shareddata She didn't say very long, in part because of on-going periodic battles with communication group trying to force her to use sna/vtam for loosely-coupled operation, as well as very little uptake (at the time) ... except for IMS hot-standby. Around the turn of the century, we would periodically drop in on the person that ran large financial transaction operation (33 liberty st, nyc) ... and he credited 100% uptime to 1) automated operator 2) IMS hot-standby ... he had triple replicated IMS hot-standby operation at geographic separated sites. slight topic drift ... when Jim Gray left IBM Research for Tandem, he palmed off bunch off stuff on me ... DBMS consulting with the IMS group, interfacing with BofA, early adopter of original relational/SQL implementation, etc ... http://www.garlic.com/~lynn/submain.html#systemr At tandem he did study of what was causing outages. One of the things he found was that hardware reliability was getting to point where it was responsible for decreasing percentage of outages and other factors were starting to dominate ... software faults, people mistakes, environmental issues like power outages, floods, earthquakes, etc). summary/overview http://www.garlic.com/~lynn/grayft84.pdf later we were doing IBM's (RS/6000) HA/CMP ... some past posts http://www.garlic.com/~lynn/subtopic.html#hacmp and working on both commercial, DBMS ... old reference to Jan1992 meeting in Ellison's conference room http://www.garlic.com/~lynn/95.html#13 as well as technical with gov. agencies and national labs ... some old email http://www.garlic.com/~lynn/lhwemail.html#medusa While out marketing HA/CMP, I coined the terms geographic survivability and disaster survivability to differentiate from disaster/recovery ... some past posts http://www.garlic.com/~lynn/submain.html#available On the commercial side, the mainframe DB2 group were complaining if I was allowed to continue ... it would be at least five years ahead of them. Shortly later, the cluster scaleup part was transferred and announced as the IBM supercomputer for technical and scientific *ONLY* (and we were told we couldn't work on anything with more than four processors). -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: New Line vs. Line Feed
john.archie.mck...@gmail.com (John McKown) writes: As a side note (as I have heard it), the reason that Windows uses CRLF as a line ending is because MS-DOS did the same. MS-DOS used CRLF because CPM-80 used CRLF. And, finally, CPM-80 used CRLF because the common printers at the time could not do a carriage return / line feed in a single operation. So, Gary Kindall (author of CPM-80) decided to end text files with CRLF so that he didn't need to complicate the printer driver to put a LF in when a CR was detected. This made good sense in the day that 64K RAM and a 1 Mhz 8080 was top of the line equipment for the hobbyist. a little other topic drift from recent IBM antitrust thread Other trivia ... also at the scientific center ... GML was invented at the science center in 1969 (G, M, L are the 1st letters of the inventor's last name). This is posting by Sowa about GML being used by IBM for documents used in the antitrust suit http://ontolog.cim3.net/forum/ontolog-forum/2012-04/msg00058.html from above: For text that was copied from the original OED, they got GML to produce exactly the same line breaks and hyphenation. They needed to get it exactly right in order to aid the proof readers who had to make sure that the new copy was identical to the old. The GML-based software in the 1980s was far more flexible than MS Word is today. Just look at the OED and imagine how you might use MS Word to match that exactly. ... snip ... in the mid-60s at science center, CMS script was implementation of CTSS runoff using dot formating controls ... then later, script was enhanced to support GML tag processing. in late 70s, a vm370 SE in the LA branch ... did implementation of CMS script on trs80 (NewScript) and periodically mentioned ... before ms/dos http://en.wikipedia.org/wiki/MS-DOS there was seattle computer http://en.wikipedia.org/wiki/Seattle_Computer_Products before seattle computer there was cp/m, http://en.wikipedia.org/wiki/CP/M before cp/m, kildall worked with cp67/cms at npg http://en.wikipedia.org/wiki/Naval_Postgraduate_School other Sowa trivia ... on the failure of FS and how poorly 3081 compared to competition http://www.jfsowa.com/computer/memo125.htm -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: New Line vs. Line Feed
t...@vse2pdf.com (Tony Thigpen) writes: It's actually much worse. There are three: Ebcdic: CR = x0D NL = x15 LF = x25 Originally, CR only moved the print back to the first position of the same line. LF only moved the print down one line without moving sideways. NL moved both down and to the first position of the line. When it was designed, they were using teletype machines and simple printers. No CRTs. Historically: 1930's had the Teletype standard: International Telegraph Alphabet No. 2 (ITA2); which had both a CR and a LF and required both at the end of a line. 1950's IBM introduces BCD and adds NL 1960's IBM introduces EBCDIC and continued using the 3 values. 1960's ATT pushes for a replacement of ITA2 which the ATA published as ASCII in 1963. (One of their requirements was 7 bit so EBCDIC was ruled out.) In the ASCII world, CR and LF were the standard until the mid-1960's when the Multics developers decided that using two characters was stupid and they started using just LF. Unix and follow-on OSs carried on the same tradition. Today, it's a mess. Windows wants CRLF. Internet RFCs normally use CRLF. Mac and Linux use just LF. Interesting, Windows Notepad requires CRLF, but Windows Wordpad will read and display a LF only file correctly and even change the file to CRLF when saved. IBM did much of the standardization for ASCII and 360 originally was suppose to be an ASCII machine ... unfortunately the 360 ASCII unit record gear wasn't ready ... and the decision was made to go (temporarily) with the old BCD unit record gear (but there was some unfortunate side-effects of that decision). EBCDIC and the P-Bit, The Biggest Computer Goof Ever http://www.bobbemer.com/P-BIT.HTM The culprit was T. Vincent Learson. The only thing for his defense is that he had no idea of what he had done. It was when he was an IBM Vice President, prior to tenure as Chairman of the Board, those lofty positions where you believe that, if you order it done, it actually will be done. I've mentioned this fiasco elsewhere. ... snip ... by the father of ASCII http://www.bobbemer.com/FATHEROF.HTM his history index http://www.bobbemer.com/HISTORY.HTM -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: PCI DSS compliance for z/OS
charl...@mcn.org (Charles Mills) writes: I think much of the problem is with credit card numbers themselves. There are only ~10**16 possible credit card numbers -- many fewer if you allow for the fact that only certain combinations are valid. A credit card number is easier to brute-force guess than its encryption key, format-preserving or not. trivia/background long ago and far away we got called as consultants to small client/server startup that wanted to do payment transactions on their server, they had also invented this technology called SSL they wanted to use; the result is now sometimes called electronic commerce. Then somewhat having done electronic commerce, in the mid-90s we were asked to participate in the (financial industry) x9a10 standards working group which had been given the requirement to preserve the integrity of the financial infrastructure for all retail payments (not just internet, *ALL*, point-of-sale, attended, unattended, credit, debit, ACH, i.e. *ALL*). part of this is we did end-to-end threat analysis and attetmpted to use a number of metaphors to characterize the existing paradigm: dual-use ... since information from previous transactions can be used for fraudulent transactions, that information has to be kept totally confidential and never divulged. at the same time the same information is required in dozens of business processes at millions of locations around the world. we've periodically commented that even if the planet was buried under miles of information hiding encryption, it still wouldn't stop leakage security proportional to risk ... the value of the transaction information to the merchants is the profit on the transactions, which can be a couple dollars (and a couple cents for the transaction processor) ... the value of the information to the crooks is the account balance and/or credit limit ... as a result the crooks can afford to outspend the defenders by a factor of 100 times. ... the x9a10 financial standard working group came up with a transaction standard that eliminated the dual-use characteristic of the account number ... which met that it no longer needed to be kept hidden (and the earlier work we did on electronic commerce was the major use of SSL for hiding the account number ... which in the new transaction standard was no longer necessary). Later we were tangentially involved in the cal. state data breach notification law. A lot of the participants were heavily involved in privacy issues and had done detailed, in-depth public surveys. The #1 issue was identity theft, primarily of the form of fraudulent financial transactions as the result of breaches and there was little or nothing being done about the breaches. An issue is typically an entity/institution takes security measures in self protection, In the case of the breaches, the institution wasn't at risk ... it was their customers. It was hoped that the publicity from the breach notifications would prompt breach countermeasures. Note in the years since the cal. state breach notification act there have been numerous federal (state preemption) acts introduced ... about evenly divided between those similar to the cal. act and those that would effectively eliminate any requirement for notification (frequently ingeniously disguised as criteria on when notification was required). The PCI DSS specification came out after the appearance of the cal. state data breach notification and referenced by federal legislation attempting to eliminate notification requirements ... because the industry was addressing the problem. Early jokes about the PCI DSS certification was that it was relatively straight-forward ... but everybody with PCI DSS certification that had a breach would have their certification revoked. -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Knowledge Center Outage May 3rd
jerry.whitteri...@safeway.com (Jerry Whitteridge) writes: I miss HONE ! Jerry Whitteridge Lead Systems Engineer Safeway Inc. I was recently asked when HONE actually shutdown http://www.garlic.com/~lynn/2015c.html#93 HONE Shutdown and found an email from may1998 saying it was going away HONE (hands-on network environemnt), some past posts http://www.garlic.com/~lynn/subtopic.html#hone had started out after the 23jun1969 unbundling announcement (starting to charge for application software, se services, etc), some past posts http://www.garlic.com/~lynn/submain.html#unbundle with (virtual machine) CP67 (running on 360/67), to give branch SEs hands-on practice with operating systems. Previously SEs got sort of journeyman training as part of large group onsite in customer datacenter, after unbundling, nobody could figure out how not to charge customers for this SE time onsite at customer. Science center very early did enhancements to CP67 that provided the simulation of the new 370 (before virtual memory instructions), so they could work with latest operating systems gen'ed for 370. For CP67/CMS, the science center ... some past posts http://www.garlic.com/~lynn/subtopic.html#545tech had also ported apl\360 to CMS for cms\apl. HONE then started offering apl-based salesmarketing support tools ... which soon came to dominate all HONE activity and the virtual guest operating system use disappeared ... and HONE clone systems would start sprouting up all over the world HONE eventually migrated to VM370 from the custom science center cp67/cms. This is old email from 40yrs ago today (30apr1975) http://www.garlic.com/~lynn/2006w.html#email750430 where I've moved a bunch of enhancements from CP67 to VM370 and made it (csc/vm) available to internal datacenters ... including HONE ... which would run my enhanced custom operating systems for another decade (including after I moved to SJR and called in sjr/vm, I then started moving off doing mainframe work). Not long after the above email, US HONE consolidated all its (US) datacenters in Palo Alto. By the end of the 70s, the US HONE datacenter was the largest single system image operation in the world ... several large (POK) multiprocessor mainframes operating in loosely-coupled operation with load balancing and workload fall-over (in case of failure ... effectively peer-coupled shared data architecture mentioned in this recent post http://www.garlic.com/~lynn/2015c.html#112 JES2 as primary with JES3 as a secondary but was not released to customers. Then in the early 80s, the US HONE complex was replicated first in Dallas and then a 3rd in Boulder ... with load balancing and fallover ... countermeasure to disaster scenarios (like earthquake in california) ... also not released to customers. for other drift, the previous post reference to IMS hot-standby ... had a fall-over operational problem. IMS configuration would be large CEC (one or more shared-memory processor) with fall-over hot-standby. IMS could immediately fall-over but VTAM sessions were enormous problem ... large systems could have 30,000-60,000 terminals ... which could take VTAM one or more hrs to get backup and running. We actually did some work with a 37x5/NCP emulator that spoofed mainframe VTAM that sessions were being managed cross-domain ... but was actually done by outboard non-VTAM processing ... which could manage replicated shadow sessions to the hot-standby machine ... so IMS hot-standby fall-over would be nearly immediately. However, this met enormous amount of resistence from the communication group (for lots of reasons, no communication group hardware and SNA RUs were carried over real network with lots of feature/function that couldn't be done in SNA). This sort of issue then in internet server environment where operations are connectionless ... in theory doesn't require long-term session maintenance overhead server workload is propotional to the workload ... not to the number of clients ... much easier to have replicated servers, load balancing and workload fall-over. The browser HTTP people did mess up ... they did use (session) TCP (instead of UDP) to implement a connectionless protocol ... it would go to all the overhead of setting up a session to do a connectionless operation and then immediately tear it down. Besides all the (unecessary) processing overhead (for TCP session setup/shutdown), TCP protocol (chatter) has a minimum of seven packet exchange. This was initially noted as webserver load started to scaleup. Industry standard TCP had what was called FINWAIT (to handle session dangling packets after session was closed) list ... and ran the list linearly looking if incoming packet was part of recently closed session ... aka expectation had been FINWAIT list was empty or a few entries. Increasing HTTP webserver workload started to have thousands of entries on the FINWAIT list ... and FINWAIT processing would consume 95% of webserver processor. We had been
Re: JES2 as primary with JES3 as a secondary
sipp...@sg.ibm.com (Timothy Sipples) writes: Let's take a brief look at this not exactly new history. I can fairly easily trace JES3 back a quarter century. (Perhaps somebody else would like to go back into the pre-Sysplex JES3 era, from 1973 to 1990, to see what IBM recommended and/or required.) trivia ... my wife was in the gburg JES group and part of the catchers of ASP (from the west coast) to turn into JES3. She then was part of the authors of JESUS (JES unified system) that combined all the features that the JES2 and JES3 customers couldn't live w/o ... but it never got very far because of various internal political issues. She then got con'ed into going to POK to be in charge of mainframe loosely-coupled architecture ... where she did peer-coupled shared data architecture. she didn't remain long because her architecture1 saw very little uptake ... except for IMS hot-standby (until SYSPLEX and parallel SYSPLEX). She was also being badgered by the communication group to force her into using SNA for loosely-coupled operation (there would periodically be a truce where communication group had strategic ownership of everything that crossed the datacenter walls and she could use whatever she wanted within a datacenter ... but then they would start badgering again). I can't speak to the other issues ... but on the JES2 networking side in the 70s 80s ... not only couldn't JES2 talk to anything else ... talking to another JES2 at a different release level could result in taking down both JES2 and the MVS system. The issue was that JES2 networking implementation intermixed networking control and job control fields and minor release-to-release changes resulted in incompatible systems. On the internal network, JES nodes were kept at edge boundary nodes. Major internal network talked to JES nodes by using drivers that emulated JES protocol ... and because of the issues with JES incompatible release vulnerabilities ... a large library of internal network software drivers grew up that would not only format fields expected for specific JES release being talked to ... but also handle JES release reformating ... allowing different JES systems to communicate. I've periodically commented on the infamous case of files from San Jose disk plant site JES system resulting in Hursley MVS crashes ... and it was blamed on the Hursley internal network software. The actual issue was some new release-to-release JES field incompatibility and the internal network software driver library hadn't been updated to handle the new case (as part of countermeasure for keeping JES systems at different release levels from crashing MVS system). -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: A New Performance Model ?
sipp...@sg.ibm.com (Timothy Sipples) writes: Storage isn't what it was in 1982, and that's the whole point. It's faster, more reliable, and ridiculously less expensive. We shift our attentions elsewhere, rightly so, at least in terms of degree of emphasis. We simply don't worry about kilobytes if we're rational. This year we worry about terabytes, and maybe in the future we won't even worry about those. re: http://www.garlic.com/~lynn/2015c.html#65 A New Performance Model ? I've periodically mentioned that when measured in number of CPU cycles access to storage (aka a cache miss) is similar to 60s access to disk when measured in 60s CPU cycles (caches are the new storage and storage is the new disk). for decades other processors (especially risc and then i86 when they moved to risc cores with hardware layer that translated from i86 to risc microops), have had lots of hardware features that attempt to mitigage/compensate for (cache miss) storage access latency; hyperthreading, out-of-order execution, branch prediction and speculative execution. the claim is that at least half the z10-z196 per processor throughput improvement is starting to introduce similar features ... with further refinements moving to z12 z13. go back over 40 years, this shows up in 195. I've periodically mentioned getting con'ed into helping with effort to add hyperthreading to 370/195 ... which never announced/shipped. The issue was that 195 pipeline had out-of-order execution but didn't have branch prediction or speculative execution ... so conditional branches drained the pipeline. It took careful programming to get sustained 10MIPs throughput ... but most codes (with conditional branches) ran at 5MIPs. The objective with hyperthreading was to emulate two-processor multiprocessor hoping that two instruction streams running at 5MIPs each would archieve 10MIPs throughput. it was basically red/blue mentioned in this 60s ACS/END reference http://people.cs.clemson.edu/~mark/acs_end.html Note that the above also points out that ACS-360 was shutdown because executives thought that it would advance the state-of-the-art to fast and they would loose control of the market. It lists some of the ACS-360 features that show up more than 20yrs later with ES/9000. The equivalent to 195 pipeline careful programming ... is careful code ordering to minimize cache misses (in much the same way that 70s/80s code was ordered to minimize page faults ... requiring disk accesses). Recent discussion in comp.arch about (virtual memory and) VS/Repack out of the science center in the 70s ... which did semi-automated code reorganization for virtual memory operation. Before it was released to customers, many internal development groups had been using it for improving operation for virtual memory environment; they also used some of the VS/Repack technology for hot-spot analysis. http://www.garlic.com/~lynn/2015c.html#66 Messing Up the System/360 aka part of the decision to migrate all 370s to virtual memory. Old post that the primary motivation for this was analysis that because MVT storage management was so bad ... that regions had to be four times larger than normally used ... a typical 1mbyte storage 370/165 ran with four regions. With virtual memory, it would be possible to run with 16 regions and still result in little or no paging. http://www.garlic.com/~lynn/2011d.html#73 Multiple Virtual Memory topic drift ... what 370/195 didn't account for was that MVT/SVS/MVS in the period introduced extraordinarily inefficient multiprocessor overhead, typical guidelines was two processor operation was 1.3-1.5 the throughput of single processor. this brings up the story about compareswap ... invented by charlie at the science center when he was doing work on fine-grain (efficient) multiprocessor locking for (virtual machine) cp/67. initial attempt to have it included in 370 was rejected ... the 370 architecture owners said that the POK favorite son operating system people were claiming that testset was more than sufficient for multiprocessor support (partially accounting for their being able to only get 1.3times the throughput). cp67 ( later vm370) multiprocessor support could get close to multiprocessor hardware throughput (with minimal introduced multiprocessor operating system overhead). we ere finally able to justify compareswap for 370 with the examples of how multithreaded applications could use compareswap (regardless of single processor or multi-processor operation) ... examples that continue to be included in POO. past multiprocessor /or compareswap posts http://www.garlic.com/~lynn/subtopic.html#smp past science center posts http://www.garlic.com/~lynn/subtopic.html#545tech -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message:
Re: A New Performance Model ?
idfzos...@gmail.com (Scott Ford) writes: Agree you 100%. Maybe they need a second pair of eyes to review the design. I know I do and I will bet other software designers and system programmers do. A second pair of eyes is like a Dr.'s second option.. Like you mentioned something was missed and the easy out was a mainframe upgrade. I agree with everyone on this one, sometimes it's lack of experience too. the IBM science center pioneered a lot of performance methodologies in the 60s 70s ... hot-spot monitoring, system modeling, multiple regression analysis, etc. some of the system modeling work eventually evolves into capacity planning. One of the system models was analytical model done in APL. The APL model evolves into the Performance Predictor available on the world-wide salesmarketing support HONE system ... branch office could obtain customer workload and system profile data ... feed it into the Performance Predictor and ask what-if questions (aka what happens if the workload changes, system configuration changes, more disks, more memory, etc major objective justifying selling more hardware) Around the start of the century I ran into consultant that was making a living from performance consulting to large mainframe datacenters in Europe and the US. IBM's downturn in the early 90s, IBM was unloading some amount of its stuff ... and this consultant obtained the right to a descendent of the performance predictor and ran it through an APL-C language converter. We met at a large datacenter that had a 450kloc cobol program that ran evernight on 40+ max. configured mainframes (constantly being upgraded, none older than 18months, number required for application to finish in the overnight batch window). They application had a few dozen people in peformance department that had been working on it for decades ... primarily using hot-spot methodology. Hot-spot tends to shine light on sections that need logic examination for doing things better ... working primarily with logic at the micro-level The modeling work fed workload system activity data and identified areas that resulted in 7% improvement. I then used multiple regression analysis with application activity data to spotlight some macro-level logic that resulted in 14% improvement. Remember that this is an application that had dedicated performance group with dozens of people that had been working with this application for decades (but primarily using hot-spot methodology ... that tends to focus on micro-level logic) -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: OT: Digital? Cloud? Modern And Cost-Effective? Surprise! It's The Mainframe - Forbes
Robert Wessel robertwess...@yahoo.com writes: IBM shipped about 20 360/91s, then a couple of 360/95s with a redesigned memory subsystem, then the 360/195 which re-implemented the same machine on a faster, denser logic process, then that modified was to include the basic S/370 extensions (no virtual memory) and shipped as the 370/195. About 40 machines of all four types (combined) were shipped. in the 70s, the 195 people sucked me into doing some stuff with them on 370/195 multiprocessor that never shipped ... basically red/blue multithreading mentioned here http://people.cs.clemson.edu/~mark/acs_end.html the above also includes some other discussion of 195 ... although primarily '60s 360 ACS ... which got canceled because executives thought that it would advance the state-of-the-art too fast and they would loose control of the market ... aka acs/360 would be significantly more cost-effective machines (also describes some of the ACS features that eventually show up in the 1990 ES/9000) one of the things they told me was that another difference between 360/195 370/195 (besides the non-virtual memory 370 instructions) was hardware instruction retry ... which greatly improved reliability. 195 execution units would do 10mips but required careful programming for the pipeline ... which did out-of-order execution ... but not branch preduction or speculative execution ... so conditional branches would drain the pipeline. as a result, most codes ran around 5mips. motivation for red/blue multitreading was the 10mips execution units would be kept busy by two 5mip threads. recent posts mentioning 370/195 http://www.garlic.com/~lynn/2015.html#27 Webcasts - New Technology for System z http://www.garlic.com/~lynn/2015b.html#61 ou sont les VAXen d'antan, was Variable-Length Instructions that aren't this describes decision to make all 370 machines virtual memory ... basically MVT virtual memory allocation was so bad that typical region size had to be four times larger than what was being used ... a 1mbyte 370/165 running four regions could get 16 regions with virtual memory and still have little or no paging. http://www.garlic.com/~lynn/2011d.html#73 Multiple Virtual Memory however, retrofitting 370 virtual memory hardware to 370/165 (for 165-II) was no trivial task ... eventually they decide to drop several 370 virtual memory features because they were too hard for the 165 ... other machines would also have to drop those features ... and software groups that had already written code using the dropped features would have to be reworked. -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: New Principles of Operation (and Vector Facility for z/Architecture)
sipp...@sg.ibm.com (Timothy Sipples) writes: The IBM z13's ~139 SIMD instructions are different and new, yes. I expect that they represent a perfect functional superset of the long ago discontinued S/390 Vector Facility. However, it's probably not particularly useful to draw many parallels (!) with that older product. Yes, they are very different. As one example, every IBM z13 processor core incorporates the new SIMD instructions as a standard included feature. That's a much different, much lower latency design than the old, optional S/390 Vector Facility. If you have older code that was able to exploit the S/390 Vector Facility, I expect you could adapt it to exploit the new SIMD instructions. IBM's latest compilers can often help. However, you can do much, much more with the new instructions. Please see my other post about the IBM MASS and ATLAS libraries, for example. This IBM redpiece introduction is also a good, quick read: the 3090 processor engineers complained some about adding vector to 3090. their claim was big part of vector was that floating point processing was so slow ... that the typical memory bus utilization was very low ... as a result it was possible to have large number of floating point execution units running concurrently and still not saturate the memory bus. they claimed that they had improved 3090 floating point processing ... so that scalar floating point was capable of keeping memory bus busy. they felt that adding vector to 3090 was pure marketing (since most applications would saturate memory bus just doing scalar floating point ... and adding additional concurrent floating point execution units would rarely increase effective throughput). these days the massive supercomputers have both the data and the execution split across tens of thousands of systems. SIMD greatly expanded type of things being done http://en.wikipedia.org/wiki/Streaming_SIMD_Extensions part of this is that number of chip transisters have exploded and they are constantly looking for what they can do with all those transisters (other than design complexity, little incremental cost ... even that is mitigated with standard chip design libraries) -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Economics of Mainframe Technology
arthur.gutow...@gm.com (Art Gutowski) writes: If my notes are accurate from Ross' Keynote address to SHARE attendees in Seattle, mainframes account for 68% of production workloads, but only 6% of IT spend (exclusive of aggregate labor costs across platforms). Given the armies of sysadmins to support *nix and windoze platforms, I gotta believe labor costs on these platforms eclipse those of the mainframe. at industry level ... one of the industries that didn't migrate off mainframe was financial ... which tends to have much higher profit margin than others. in the 90s, there was big effort in the financial industry to migrate to killer micros as lots of other industries were doing ... that failed. those failures had much higher consequences in financial ... and so they've tended to retrench and minimize their risks for some period. at datacenter level ... a large cloud megadatacenter will have hundreds of thousands of systems (more processoring than the aggregate of all mainframes in the world today), massively automated with staff of 80-120 people. large cloud operators have claimed for a decade or more that they assemble their own systems for 1/3rd the cost of brand name vendors ... along with news about server chip vendors starting to ship more chips to cloud operations than to brand name vendors (possibly motivation for IBM to sell off their server chip business). there have been rumors that some of the brand name server vendors have been doing side cloud business ... for large volume order they will price close to that of the costs that the large cloud operators claim (where the massive automation is migrating out into corporations running their own clouds). -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: bloomberg article on ASG and Chpater 11
0047540adefe-dmarc-requ...@listserv.ua.edu (Bill Johnson) writes: Bankruptcies are rarely a good thing. I've been through one. trivia ... stockman goes into some detail about stock buybacks (including IBM's) and characterizes them as mini-form of LBO. http://www.amazon.com/Great-Deformation-Corruption-Capitalism-ebook/dp/B00B3M3UK6/ because of bad rep from the SL crisis, the industry changed its name to private equity and junk bonds become high-yield bonds. industry has been borrowing money for LBO and has been characterized as similar to house flipping. Difference is that the loan goes on the bought company's books and goes with it after flipping (rather than paid off; private equity can sell for less than they paid and still walk away with boat loads of money ... totally aside from what they loot might from the company). the enormous (LBO) debt load has over half corporate defaults involving companies currently or formally in private equity clutches. ... ref http://www.nytimes.com/2009/10/05/business/economy/05simmons.html?_r=0 AMEX was in competition with KKR for LBO of RJR ... and KKR wins. the president of AMEX had been in competition to be the next CEO and wins. However, KKR is then having some problems with RJR and hires the president of AMEX away to turn it around. Then IBM has gone into the red and is in the process of being broken up into the 13 baby blues. The board hires the former AMEX president away to resurrect IBM and reverse the breakup ... and then begins to apply some of the same techniques used at RJR (also start to see big upswing in stock buybacks) http://www.ibmemployee.com/RetirementHeist.shtml -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Anthem Healthcare Hacked
hal9...@panix.com (Robert A. Rosenberg) writes: What is done with the Sensitive Data is importance. In many cases, such as passwords, there is no need to know the actual data but only to compare it with some supplied value to see that it matches. Thus a stored one-way hashed value is secured since there is no way to unhash it since all that is needed is to hash the value you think it is and compare the two hashes. re: http://www.garlic.com/~lynn/2015.html#96 Anthem Healthcare Hacked an issue is something you know shared secrets for authentication, pins, passwords, as well other information you might know that can be used for authentication, mother's maiden name, social security number, date-of-birth, etc ... ... but skimming attacks can occur in the infrastructure before the data is hashed. also hashing doesn't work if working with human operators that are doing purely visual compare. one of the worst is financial industry ... where the account number tends to be dual-use ... essentially both for authentication, but also required in dozen of business processes at millions of locations around the planet (security requirements that authentication info is kept totally confidential and *NEVER* divulged, conflicting requirements when same information is also required for large number of business processes) ... harvesting can be breaches at backends, at any of the business processes, any of the transmission points and at the originating front-ends. hash for password repositories has been used for some time ... storing hashed password first done in unix in early 70s: http://en.wikipedia.org/wiki/Password#History_of_passwords trivia ... above also mentions CTSS: http://en.wikipedia.org/wiki/Compatible_Time-Sharing_System then some of the CTSS people go to the science center on 4th flr 545 tech sq ... some past posts http://www.garlic.com/~lynn/subtopic.html#545tech others go to the 5th flr and do Multics http://en.wikipedia.org/wiki/Multics some of the people working on Multics return home and do simplified version that they call Unix. http://en.wikipedia.org/wiki/Unix#History above also references Greg Chesson ... who I worked with in the 80s, when I was on the XTP technical advisory board. -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
ancient cobol applications
local news just had item about ancient software at state agencies, 619 major cobol applications developed in 80s ... frequent crashesoutages, almost impossible to maintain or change ... in part because of the lack of cobol programmers. The state is even considering setting up financial incentive for schools to produce cobol programmers. -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Anthem Healthcare Hacked
Anthem's stolen customer data not encrypted; But under federal law, health insurance companies don't have to encrypt user data. http://www.cnet.com/news/anthems-hacked-customer-data-was-not-encrypted/ In early part of century, I was co-author of financial industry privacy standard ... and we had some meetings with gov. employees that had drafted the original HIPAA legislation back in the 70s. They mentioned that special interests had kept it from being passed for decades ... and even once it was passed, there were no provisions for actually doing anything/security about it. we were also tangentially involved in the cal. state data breach legislation ... having been brought in to help wordsmith the cal. state electronic signature act. A lot of the participants were heavily involved in privacy issues and had done detailed, in-depth public surveys. The #1 issue was identity theft, primarily of the form of fraudulent financial transactions as the result of breaches and there was little or nothing being done about the breaches. An issue is typically an entity/institution takes security measures in self protection, In the case of the breaches, the institution wasn't at risk ... it was their customers. It was hoped that the publicity from the breach notifications would prompt breach countermeasures. -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: a bit of hope? What was old is new again.
shmuel+ibm-m...@patriot.net (Shmuel Metz , Seymour J.) writes: FSVO this. IBM distributed service with preassembled modules. Only if you had updates would the service process reassemble. re: http://www.garlic.com/~lynn/2015.html#84 a bit of hope? What was old is new again http://www.garlic.com/~lynn/2015.html#85 a bit of hope? What was old is new again http://www.garlic.com/~lynn/2015.html#86 a bit of hope? What was old is new again http://www.garlic.com/~lynn/2015.html#87 a bit of hope? What was old is new again vm370 monthly service tapes (referred to as PLC or program level change) had both the full original release source and all the accumulated source updates ... besides having precompiled modules. The burlington group had done a modified CMS TAPE program for release and monthly PLC service tapes (VMFPLC). Among the things lost with the shutdown of the Burlington development group was the source changes for VMFPLC (one of the few things where the full source wasn't shipped). I was possibly the only person in the company that still had the original source for VMFPLC. some discussion in this post http://www.garlic.com/~lynn/2003b.html#42 VMFPLC2 tape format In the late 70s, I was doing an internal backup/archive for internal datacenters and enhanced the VMFPLC source to add some additional features and get higher tape data capacity (lot more tape record blocking so there were fewer physical tape records). Originally distributed mostly in the silicon valley area (including HONE) ... but started to spread through much of the rest of the company. This went through several internal releases ... and then was enhanced for customer release with lots of client applications for backing up distributed environment and released as workstation datasave ... which morphs into ADSM ... and then when IBM was unloading the disk division, morphs int TSM. some old email http://www.garlic.com/~lynn/lhwemail.html#cmsback and past posts http://www.garlic.com/~lynn/submain.html#cmsback -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: a bit of hope? What was old is new again.
edgould1...@comcast.net (Ed Gould) writes: yet IBM never delivered a source code maintenance system. Something that practically everyone was in need of. re: http://www.garlic.com/~lynn/2015.html#84 a bit of hope? What was old is new again. science center did the multi-level cms update source maintenance system as part of joint project with endicott to implement cp67 370 virtual machine emulation on 360/67. the non-virtual memory 370 emulation was originally used by branch office HONE online cp67 systems to test new operating systems. the full virtual memory 370 emulation was used for development of 370 virtual memory operating systems (i.e. 360/67 cp67 virtual memory machine emulation was in regular production use a year before 370 virtual memory hardware became available). cp67 distribution was always full source ... and customers typically built their systems from the source. This continued with the vm370 followon ... new releases had single source file per module ... and then monthly maintenanceenhancement distributions was done via incremental add-on updates ... with cumulative source included updates included on every monthly maintenanceenhancement distribution. New releases would merge the incremental updates into base source file and things would start again ... accumulating increasing number of incremental source updates. SHARE waterloo updates and customers used the same process for their source changes ... and large part of internal development did also (which accounted for the origin of a lot of the OCO-wars). Not that new releases ... besides past incremental updates being incorporated into base source files ... there could also be large amount of new function/code added ... never before seen by customers as incremental updates ... increasing the difficulty of release to release migration. Tools were developed (both inside at customers) that would analyze new source releases and pick out differences from the latest previous release (with all maintenance/changes applied) to facilate the release to release source transition. There were also periodic internal fights ... where various MVS-based product (like JES2) did all their stuff with CMS source maintenance ... but were required to convert to the official internal MVS process for final integration. Note after FS failure http://www.garlic.com/~lynn/submain.html#futuresys qd 3033 (starting out as 168-3 logic mapped to 20% faster chips) and 3081 efforts were kicked off in parallel as part of mad rush to get stuff into the 370 product pipelines. during the 3033 product life, there started to be minor (supervisor state) tweaks made to the machine which were mandatory for new operating system releases. The clone makers initially responded with operating system patches to work with non-tweak hardware. As patching the operating system was made more and more difficult, the clone makers eventually responsed with macrocode ... basically 370 instructions running in new machine mode that would implement the tweak features this enormously simplified the implementation of such features ... compared to the enormous difficulty involved in generating native microcode. This shows up in 3090 timeframe when clone vendor has used macrocode to create hypervisor support ... and it was a much larger ( longer) effort for 3090 to eventually respond with PR/SM. In the current timeframe, things could be construed as customers having their own programming support staff represents money that could otherwise be spent on vendor softwareservices (2012 claim that processor sales represented 4% of revenue ... but total mainframe group, including softwareservices was 25% of total revenue and 40% of profit). The same efforts to inhibit clone vendor patches ... also increasingly made it difficult for customers to move their changes to new releases (they either stayed on their old hardware or moved to new clone hardware that worked with the older releases). The OCO-wars could be viewed as both inhibiting new operating system versions working on clone processors and minimizing customer migration latency to latest software releases and hardware models. One of the worst case examples starts during the FS period, I continued to work on 370s (and periodically ridicule FS). Also one of my hobbies was producing highly enhanced production operating system distribution for internal datacenters (science center was on 4th flr of 545 tech sq, and multics was on 5th flr of 545 tech sq, at one point I would needle the multics crowd that I had more internal datacenters running my enhanced operating systems than all the datacenters in the world running multics). Anyway, for some reason, one of these versions was made available ATT longlines ... which then made a lot of their own enhancements and distributed it throughout a lot of ATT. Nearly a decade later the IBM ATT national sales rep tracks me down to ask me to help with ATT. The decade old operating system, ATT would
Re: a bit of hope? What was old is new again.
edgould1...@comcast.net (Ed Gould) writes: So, it was IBM saying if you don't run VM, FY? I think the many MVS sites would take exception to that. From my perspective VM was OK some things but not for PRODUCTION. VM was a sand box so the real work was to be done on MVS. re: http://www.garlic.com/~lynn/2015.html#84 a bit of hope? What was old is new again. http://www.garlic.com/~lynn/2015.html#85 a bit of hope? What was old is new again. depends on what you mean by real work. Nearly all online, network and various other kinds of real work inside IBM ... went on with VM. there was the virtual machine based (initially cp67, then vm370), world-wide online salesmarketing support HONE system ... and by mid-70s, an IBM mainframe couldn't be ordered w/o having been processed on HONE. some past HONE posts: http://www.garlic.com/~lynn/subtopic.html#hone when I first transferred from Cambridge http://www.garlic.com/~lynn/subtopic.html#545tech to San Jose Research, they let me wander around various locations in silicon valley (there use to be a joke that I worked 4shifts a week, 1st shift in SJR, 2nd shift in disk engineering, 3rd shift in STL, now SVL, and 4th shift/weekends at HONE). when I first transferred, I found the disk engineering and development lab was all being done stand-alone, mainframes being scheduled 7x24 for testing. At one point they had tried to use MVS (allowing multiple concurrent testing), but in that environment, MVS had 15min MTBF (hang/fail requiring manual restart). I offerred to rewrite I/O supervisor so that it would be bullet proof and never fail, enabling. multiple concurrent, on-demand testing ... significantly improving productivity. I wrote an internal document describing all the enhancements ... and happened to include the MVS 15min MTBF reference ... which drove the POK MVS over the edge (not that I couldn't prove it true, but that I had made the information public inside IBM) ... and I was eventually told that while they couldn't actually get me fired, but they would be able to make sure that I never got an corporate award for the work. One of the other side-effects was since it was my software, any problems disk enginneering would constantly suck me into working on their problems. They also start insisting that I sit in on conference calls with pok channel engineers. past posts mentioning getting to play disk engineer in bldgs. 1415 http://www.garlic.com/~lynn/subtopic.html#disk old email reference that 3380s about to ship and MVS system is hanging/crashing in *all* the standard FE error injection tests (and in 2/3rds of the cases, MVS leaves no indication what caused the failure). http://www.garlic.com/~lynn/2007.html#email801015 note that the internal network originated at the cambridge science center (virtual machine based) and was larger than arpanet/internet from just about the beginning until sometime in the middle 80s. Most of the nodes were vm370 and much of source development went on vm370 systems (even for mvs based products). part of the issue was severe limitation with the MVS networking support ... limit on max defined nodes were much less than total number of internal nodes ... also mixed up design so networking job control information was intermixed in headers ... resulting in traffic between dissimilar MVS releases crashing MVS. As a result, any MVS systems were restricted to edge nodes, fronted by VM370 systems that had special code that would convert all header information into exact format required by the directly connected specific release of MVS. http://www.garlic.com/~lynn/subnetwork.html#internalnet Note that this wasn't SNA ... at least not until the late 80s when the communication group insisted that it convert to SNA ... which was major factor leading to its demise. the same technology was also used for the corporate sponsored university network BITNET (EARN in europe) which was also larger than arpanet/internet for a time. some past posts http://www.garlic.com/~lynn/subnetwork.html#bitnet Also, the original relational/sql System/R was developed on 370/145 vm370 at san jose research in the 70s. STL was responsible for IMS ... but nearly all the developers did their work on vm/cms. The followon to IMS was EAGLE ... and while the corporation was preoccupied with EAGLE, it was possible to get tech. transfer for System/R to Endicott released as SQL/DS. When EAGLE finally imploded, they wanted to know how fast could System/R be ported to MVS ... eventually released as DB2. -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: a bit of hope? What was old is new again.
re: http://www.garlic.com/~lynn/2015.html#84 a bit of hope? What was old is new again. http://www.garlic.com/~lynn/2015.html#85 a bit of hope? What was old is new again. http://www.garlic.com/~lynn/2015.html#86 a bit of hope? What was old is new again. part of customer facing issue was that in the aftermath of FS failure in the mid-70s http://www.garlic.com/~lynn/submain.html#futuresys and the mad rush to get stuff back into the 370 product pipelines, the head of POK convinced corporate to kill-off vm370/cms product, shutdown the burlington mall development group and transfer all the people to POK, otherwise he wouldn't be able to ship MVS/XA on schedule in the 80s. They weren't going to tell burlington until the very last minute to try and minimize the number of people that might escape ... however the information leaked and lots of the people got away (one of the jokes was that the head of POK was one of the major contributors to DEC VAX/VMS). Endicott managed to save the vm370/cms product mission ... but had to reconstitute a development group from scratch ... and lots of the stuff that was in progress in burlington never resurfaced. also it put major damper on significant enhancements in customer releases. HONE had major enhancements for single-system-image, cluster operation with load-balancing and sharing across multiple multiprocessor systems, even being able to handle fall-over between geographically distributed complexes. A little of this finally shows shows up in customer release 30yrs later in 2009. HONE was able to accomplish these enhancements despite being under enormous pressure to move to MVS platform ... repeatedly they would be directed that they had to move to MVS ... and put all resources into the effort for a year or more ... only for it to eventually be declared a complete failure. http://www.garlic.com/~lynn/subtopic.html#hone They even tried blaming me for HONE inability to move to MVS platform ... because HONE was one of my long time internal customers for enhanced production operating systems (back to cp67 days). note that in the early 70s, CERN did a share mvs/tso - vm370/cms bakeoff report ... copies internally were classified confidential - restricted ... aka available on need to know basis only ... because it made a mockery of what POK was claiming internally (even tho it was freely available outside IBM). CERN SLAC were long time production vm370 customers ... and the first webserver outside of europe/cern was on the slacvm system. http://www.slac.stanford.edu/history/earlyweb/history.shtml first web pages http://www.slac.stanford.edu/history/earlyweb/firstpages.shtml somewhat topic drift, slac did a bit-slice 168E that implemented sufficient problem state to run 370 fortran programs ... they were used to do initial data reduction from sensors along the linear accelerator. this was then upgraded to 3081E and they were used at both SLAC and CERN for offline, initial data reduction as well as online compute farm. recent post in a.f.c. with other slac references: http://www.garlic.com/~lynn/2015.html#79 -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: a bit of hope? What was old is new again.
000433f07816-dmarc-requ...@listserv.ua.edu (Paul Gilmartin) writes: Where I read: ... For example, one thing I try to do is to have our IT infrastructure employees trained to code so that they can automate repetitive tasks. In contrast to the Enterprise mindset frequently apparent here: We don't want our general IT infrastructure employees coding. And even (though less frequently lately): How can I prevent my coders' using Unix System Services? OTOH a current thread in IBMVM: https://listserv.uark.edu/cgi-bin/wa?A2=ind1502L=IBMVMO=DF=S=P=1987 missing CUA 2001 package files... ... explores the legal, technical, and social hazards of garage tools development. IBM used FUD during the OCO-wars in the early 80s including the enormous risks of customers having source and allowing their programmers to change it. It was part of transition that included charging for operating system software. there was some study that the internal datacenters had enormous library of operating system changes ... and there was similar number of LOC in the waterloo library in the 23jun1969 unbundling announcement, that started to charge for (application) software, SE services, maintenance, etc ... the company managed to make the case that operating system software should still be free. some past posts http:///www.garlic.com/~lynn/submain.html#unbundle then during the Future System period in the first part of the 70s, internal politics was killing off 370 efforts (FS was completely different and going to completely replacing 370). The dearth of 370 products during the FS period is credited with giving 370 clone processor makers a market foothold. some past posts http://www.garlic.com/~lynn/submain.html#futuresys with the death of FS, there was a mad rush to get stuff into the 370 product pipelines ... which contributed to decision to release a bunch of (370) stuff that I had been doing all during the FS period (although I would periodically ridicule the FS activity which wasn't exactly career enhancing). Some of my stuff was then selected as guinea pig starting to charge for operating system softare (some claims part of the countermeasures to clone makers). -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: z13 new(?) characteristics from RedBook
shmuel+ibm-m...@patriot.net (Shmuel Metz , Seymour J.) writes: What is CP, chopped liver? trivia ... (at least) 80s90s ... the various vendor UNIX ports to mainframe ran under vm370 ... the issue was relying on vm370 for error handling/recovery/EREP ... because adding such capability to UNIX was several times larger effort than the straight-forward ports ... *AND* hardware field support said they wouldn't maintain system w/o it. re: http://www.garlic.com/~lynn/2015.html#43 z13 new(?) characteristics from RedBook http://www.garlic.com/~lynn/2015.html#44 z13 new(?) characteristics from RedBook http://www.garlic.com/~lynn/2015.html#45 z13 new(?) characteristics from RedBook more trivia, UTS was ATT UNIX port ... but aix370/aixesa was port of UCLA's (unix work-a-like) LOCUS (and unrelated to workstation AIX). LOCUS provided for distributed transparent operation across dissimilar architecture (aix/370 was announced in combination of aix/386) http://en.wikipedia.org/wiki/LOCUS_%28operating_system%29 -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: z13 new(?) characteristics from RedBook
000a2a8c2020-dmarc-requ...@listserv.ua.edu (Tom Marchant) writes: Today's processors have cache because main memory is _really_ slow compared to the processor. When the processor accesses something at a memory address, if the data at that location is in the cache, the processor can access it in one clock cycle (if it is in the on-chip cache) or a few clock cycles if it is farther away. i.e. current latency to access memory (on cache miss) when measured in number of processor (clock) cycles ... is compareable to 60s disk access latency when measured in terms of number of (60s) processor (clock) cycles (cache is the new memory, main memory is the new disk). that enormous idle time was what drove multiprogramming/multithreading (in the 60s). this old account that increasing the level of multiprogramming is what drove decision to move to virtual memory for all 370s ... that and the extremely inefficient MVT memory management ... aka MVT regions typically needed to be four times larger than actively used memory typical 1mbyte 360/165 MVT ran four initiators, adding virtual memory could it increase it to 16 initiators with little paging impact. http://www.garlic.com/~lynn/2011d.html#73 Multiple Virtual Memory introduction of out-of-order instruction execution for z196 is claimed to account for at least half the (per processor) performance increase from z10 to z196. this is basically a technique when one instruction is stalled waiting for memory, switch to another instruction in the same instruction stream (sort of multithreading of instruction execution at the micro level). This can only go only so far, in part because it can get quickly complex with subsequent instructions stalled waiting for results from previous instructions. recent ref (hyperthreading in the early 70s) http://www.garlic.com/~lynn/2015.html#27 Webcasts - New Technology for System z -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: z13 new(?) characteristics from RedBook
dcrayf...@gmail.com (David Crayford) writes: Better to compare it to the POWER arch http://www-03.ibm.com/systems/resources/pwrsysperf_SMT4OnP7.pdf. It may be CISC not RISC but those lines are getting more blured with every new churn of z. I would imagine that the SIMD vector units also originate from POWER. It may seem far fetched but I can see a time in the not too distant future when the two architectures are converged. note that in the 90s, i86 cisc started moving to hardware layer that translated cisc instructions into risc micro-ops for execution ... pentium pro, 20yrs ago http://en.wikipedia.org/wiki/Pentium_Pro from above: The Pentium Pro incorporated a new microarchitecture in a departure from the Pentium x86 architecture. It has a decoupled, 14-stage superpipelined architecture which used an instruction pool. The Pentium Pro (P6) featured many advanced concepts not found in the Pentium, although it wasn't the first or only x86 processor to implement them (see NexGen Nx586 or Cyrix 6x86). The Pentium Pro pipeline had extra decode stages to dynamically translate IA-32 instructions into buffered micro-operation sequences which could then be analysed, reordered, and renamed in order to detect parallelizable operations that may be issued to more than one execution unit at once. The Pentium Pro thus featured out of order execution, including speculative execution via register renaming. It also had a wider 36-bit address bus (usable by PAE), allowing it to access up to 64GB of memory. ... snip ... ... this was pipelined so wasn't serialized ... so there has been shrinking difference between popular cisc and risc for a couple decades. http://en.wikipedia.org/wiki/Instruction_pipeline above mentions early pentium4 (2000) with 20 stage pipeline and later pentium4 with 31 stage pipeline recent refs http://www.garlic.com/~lynn/2014m.html#164 Slushware http://www.garlic.com/~lynn/2014m.html#166 Slushware http://www.garlic.com/~lynn/2014m.html#170 IBM Continues To Crumble other recent posts on subject: http://www.garlic.com/~lynn/2015.html#35 [CM] IBM releases Z13 Mainframe - looks like Batman http://www.garlic.com/~lynn/2015.html#36 [CM] IBM releases Z13 Mainframe - looks like Batman http://www.garlic.com/~lynn/2015.html#38 [CM] IBM releases Z13 Mainframe - looks like Batman http://www.garlic.com/~lynn/2015.html#39 [CM] IBM releases Z13 Mainframe - looks like Batman http://www.garlic.com/~lynn/2015.html#40 [CM] IBM releases Z13 Mainframe - looks like Batman http://www.garlic.com/~lynn/2015.html#41 [CM] IBM releases Z13 Mainframe - looks like Batman http://www.garlic.com/~lynn/2015.html#42 [CM] IBM releases Z13 Mainframe - looks like Batman http://www.garlic.com/~lynn/2015.html#43 z13 new(?) characteristics from RedBook -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: z13 new(?) characteristics from RedBook
john.archie.mck...@gmail.com (John McKown) writes: Yes. I remember some decades back reading that CISC was going to die due to RISC performing better with optimizing compilers. That both did and didn't come true. The hardware exposed ISA is dominated by CISC on the high end (RISC ISA chips that I know of are ARM, Sparc, and MIPS) but the hardware internally is more like RISC. Conceptually, a bit like what IBM did with the TIMI for the i systems. Except that TIMI, from what I've read, is actually compiled into native code on the first execution and is store in a hidden portion of the executable on disk. Said compiled code is foot printed and recompiled if the TIMI object is change or, sometimes, when maintenance is applied to the i system software. I found the concept fascinating. re: http://www.garlic.com/~lynn/2015.html#44 z13 new(?) characteristics from RedBook lowmid range 360/370s were vertical microcode with standard cisc engines. in the late 70s an effort was started to move the large number of internal microprocessors to 801/risc (mostly 801/risc Iliad chips) ... including all of the 370 native engines (4361/4381 followon to 4331/4341), the as/400 (aka i-system,) and numerous controllers. For various reasons all of these efforts faultered (and you found various 801/risc engineers leaving and going to other vendors, spawning their risc projects) and reverting to business-as-usual cisc. A decade after as/400 ships with cisc, it (finally) migrated to (risc/801) power/pc. some past posts mention 801/risc, iliad, romp, rios, fort knox, power, power/pc, etc http://www.garlic.com/~lynn/subtopic.html#801 one of the earliest such efforts was the AMD 29K (and IBM may have sued AMD because former 801/risc engineer worked on it). Some (former) IBMers showed up working on HP snake risc ... and later Itanium. 801/risc as the native engine for microcoded 370 ... including some work on just-in-time (JIT) compiling ... sequences of 370 code segments dynamically translated to native risc (instead of repeated straight interpretation). Equivalent JIT was later done for some of the i86-based 370 simulators ... and analogous JIT is being done for JAVA. other trivia, AIM/Somerset (applie, ibm, motorola) for single chip 801/risc (aka power/pc) somewhat combined 801/risc rios with Motorola's 88k risc (internal IBM 801/risc had long history of not supporting cache consistency making multiprocessor implementations difficult, Motorolo's 88k did have cache consistency support). The pentium-pro translate from i86 to risc could be considered similar to (long history of) 360/370 microcoded implementations ... but at the hardware layer and pipelined. The 360/370 microcode implementations typically avg. 10 native instructions for every 360/370 instruction and since they were serialized, it needed a 10mip processor to get 1mip 360/370. The pentium-pro pipelined just needed 1mip native since it was doing several things, concurrently, in parallel with the pipeline. -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Young's Black Hat 2013 talk - was mainframe tribute song
charl...@mcn.org (Charles Mills) writes: I am not certain that MVS exposures versus lax security is a black and white dichotomy. It's easy to look after the fact at any breach and say aha! You should not have done X. I don't think the role of we security practitioners is solely pointing out exposures in MVS to IBM. I think helping customers with common less-than-ideal practices is more important. Logica was a professional service bureau with a professionally-maintained z/OS. They got breached. One might infer that other MVS sites, and not just those with lax (however defined) security practices, might also be vulnerable. long ago and far away we were brought in as consultants to small client/server startup that wanted to do payment transactions on their server; they had also invented this technology they called SSL that they wanted to use, the result is now frequently called e-commerce. early experience found that RDBMS-based ecommerce servers had more frequent exploits than flat-file based ecommerce servers ... these weren't intrinsic to the environment ... it was that RDBMS-based ecommerce servers were a lot more complicated ... and as a result people were more prone to making mistakes resulting in exploits (there is some amount of security literature about exploits proportional to complexity, which is a counter to the periodic meme of security through obscurity). much more recently there have been some SQL-specific attacks http://en.wikipedia.org/wiki/SQL_injection which claims that they can attack any type of SQL database (although a case might be made that SQL-injection is another characteristic of RDBMS/SQL being more complex). disclaimer: I periodically have stressed KISS as a major security theme. -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Slushware
alan_altm...@us.ibm.com (Alan Altmark) writes: Yet you never hear millicode being applied to storage controllers or other parts outside of the processor. And you know as well as I do that they aren't replacing microcode on the processor chips. They're replacing the OS and the applications that use them. But we continue to call it microcode. The joke's on us re: http://www.garlic.com/~lynn/2014m.html#161 Slushware http://www.garlic.com/~lynn/2014m.html#163 Slushware http://www.garlic.com/~lynn/2014m.html#164 Slushware http://www.garlic.com/~lynn/2014m.html#166 Slushware 79/80 there was effort to replace the myriad of internal microprocessors with 801/risc ... 801 Iliad chips for the low mid-range 370s, 801 ROMP chip for the follow-on to the displaywriter, new 801 chip for the AS/400 (follow-on to s/36 s/38), 801 chips for wide variety of (disk, tape, communication, etc) controlers, etc. For various reasons all of these failed and things returned to business as usual with various CISC chips ... and started to see 801 chip engineers leaving to other vendors to work on risc programs there. the followon to 4331/4341, 43614381 were originally to be 801 microprocessors with 370 simulation done in 801 software ... rather than whatever preceeding CISC processors were used (vertical microcode that avg. ten native instructions per 370 instruction). There was even work on JIT (just in time dynamic compiling of 370 into native 801/risc) ... somewhat analogous to what is seen with some modern day JAVA. I helped with white paper that shot down the use of 801/Iliad for 4381 ... the story was that CISC chips were getting sophisticated enough that much of 370 instructions could be directly implemented in silicon ... rather than having to be all simulation in microcode (software) ... resulting in significant better price/performance. as/400 eventually abandoned 801/risc implementation, changing to traditional CISC microprocessor. However, a decade later AS/400 did move over to 801/risc with power/pc. past 801/risc, iliad, romp, rios, power, etc posts http://www.garlic.com/~lynn/subtopic.html#801 A little later, IBM Germany did the (native) 370 ROMAN chipset. Somehow somebody in Nixdorf (did 370 clones) came into possession of detailed specs. for ROMAN. He sent it to somebody at Amdahl that he had been working with ... who presented it to me to return to the rightful owners (trying to avoid any litigation that might come from having come into the possession of the document). Turns out that I was trying to get a project going to package a few dozen ROMAN chipsets in a rack. It was sort of followon to something I had gotten dragged into a few years earlier. I had access to engineering 4341 (before first customer ship) and got asked to do some benchmarking for LLNL that was looking at getting 70 4341s for compute farm (sort of precursor to modern grid supercomputing). A cluster of 4341s had more computer power than high-end mainframes, were much cheaper, and required much less floor space and environmentals. old 4341 email http://www.garlic.com/~lynn/lhwemail.html#4341 later I got involved in doing something similar ... but packing as many 801/RIOS chips in a rack as possible (instead of 370/ROMAN). some old email http://www.garlic.com/~lynn/lhwemail.html#medusa -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Slushware
re: http://www.garlic.com/~lynn/2014m.html#161 Slushware http://www.garlic.com/~lynn/2014m.html#163 Slushware http://www.garlic.com/~lynn/2014m.html#164 Slushware as an aside ... the hardware layer from i86 instructions to risc micro-ops for execution ... isn't serialized ... it is pipelined operation ... simple version starts with overlapping instruction fetch decode with instruction execution http://en.wikipedia.org/wiki/Instruction_pipeline the above mentions that pentium4/pentuimD had 31-stage pipeline ... longest in mainstream consumer computing longer pipeline affects the latency for any specific instruction getting executed ... but isn't (necessarily) limiting in the aggregate instruction execution rate (since the operations are overlapped in parallel). there was recent claim (in ibm linkedin discussion) that there is approx. mainframe aggregate 18-20 milllion MIPS in the world today ... or the equivalent of around 270 max. configured EC12s (@75BIPS) ... or about 15 e5-2699v3 blades (@1.3TIPS). A typically cloud megadatacenter can have several hundred thousand blades ... and a standardized virtualization/container facility goes a long way to simplifying the operation. http://www.networkcomputing.com/cloud-infrastructure/virtual-machines-vs-containers-a-matter-of-scope/a/d-id/1269190 -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Slushware
000433f07816-dmarc-requ...@listserv.ua.edu (Paul Gilmartin) writes: It began nearly a half century ago with microcode implementation of S360 models, and only slightly later, W. M. Waite's Mobile Programming System. Nowadays: microcode-millicode-PR/SM-VM-JVM-byte code How many layers have I neglected? Hercules is a confluent branch. note that the original hypervisor was done by Amdahl in something called macrocode ... which was a layer above microcode and very close to standard 370. In the mid-70s, I had been sucked in by Endicott to help with microcode assists for 138/148 ... vertical microcode machine that avg. 10 microcode instructions per 370 instruction (not that different from the various intel based simulators). Was told that there were 6kbytes available for microcode and kernel instruction sequences dropped into microcode on nearly byte for byte ... so was to identify the top 6kbytes worth of kernel instruction sequences ... that would be moved to microcode for a 10:1 performance improvement. Old post with results of analysis ... turns out top 6kbytes of instruction sequences accounted for 79.55percent of kernel time. http://www.garlic.com/~lynn/94.html#21 In any case, I was giving presentations on the effort at the monthly bay area user group meetings (BAYBUNCH) held at SLAC ... and the Amdahl people were pumping me for additional details at the get togethers held at local watering holes after the meetings (this was before hypervisor was announced). After hypervisor was announced ... the 3090 was eventually forced to respond with PR/SM. Part of the issue was that 3090 was horizontal microcode machine ... which was enormously more difficult to program for than 370 instructions ... and was much more difficult. I had been told that Amdahl had original evolved macrocode to respond to the enormous number of architecture tweaks that IBM had been doing on their high-end (vertical microcode machines) starting with the 3033 and continued through 3081 (macrocode used to drastically reduce the effort needed to respond). I've mentioned before ... during FS period ... internal politics were killing off 370 efforts (the lack of 370 products during this period is credited with giving clone processor makers a market foothold) ... then when FS imploded there was mad rush to get 370 products back into pipeline. POK kicked off 3033 (initially 168 logic remapped to 20% faster chips) and 3081 in parallel ... more detailed account here: http://www.jfsowa.com/computer/memo125.htm since that neither 3033 or 3081 were really that competitive, the architecture tweaks would supposedly give the machines competitive advantage ... many were claimed to be performance improvements ... but folklore is that many actually ran slower (than native 370). Part of the issue is the high-end, horizontal microcode machines were profiled in terms of avg. machine cycles per 370 instruction ... by 3033, this was done to cloe to one machine cycle per 370 instruction (370 instruction move to microcode couldn't see the 10:1 improvement seen on the vertical microcode machines). In anycase, it sort of drove Amdahl into creating macrocode as a way of drastically simplifying being able to respond to the increased architecture tweaking. The other factor was that part of the mad rush after FS failure, the head of POK managed to convince corporate to kill off vm370, shutdown the development group and move all the people to POK ... or otherwise POK would be able to make mvs/xa ship schedule several years (Endicott managed to save the vm370 product mission, but had to reconstitute a development group from scratch). Part of the POK activity was creating a XA vritual machine VMTOOL (to support MVS/XA development) that was never intended to be made available to customers. After initial introduction of 370/xa and MVS/XA ... there was very slow uptake ... customers continued to run 3081s in 370 mode with MVS (or vm370). The decision then was to release the VMTOOL as the migration aid ... allowing customers to run both MVS and MVS/XA concurrently on the same machine as aid to migration. Amdahl solution was the hypervisor which provided the same capability ... but much more efficiently. IBM eventually responded with PR/SM on the 3090 ... but it was much greater effort because it required being all done in native horizontal microcode. The POK(/Kingston) group then pushed very hard to have migration aid morph into standard VM/XA product. The problem was that VMTOOL had only been developed for MVS/XA development and lacked lots of function and performance features (especially compared to vm370 of the period) and was going to require lots of resources, effort and time to bring up to compareable level of vm370. Somebody at an internal datacenter had made the changes to vm370 to provide full function 370/XA support ... which would have been trivial to release. In the internal politics between POK and Endicott, POK managed to prevail and
Re: Slushware
000433f07816-dmarc-requ...@listserv.ua.edu (Paul Gilmartin) writes: How many layers have I neglected? Hercules is a confluent branch. re: http://www.garlic.com/~lynn/2014m.html#161 Slushware http://www.garlic.com/~lynn/2014m.html#163 Slushware for other hercules drift ... risc processors had performance advantage over intel ... risc having made extensive use of technolology to compensate for the increasing mismatch between memory latency and processor speed ... out-of-order execution, speculative execution, branch-prediction, etc ... sort of the hardware equivalent of '60s multiprogramming to keep processor busy while waiting for disk access (current memory latency, measured in count of cpu cycles is compareable to 60s disk latency when measured in number of 60s cpu cycles). however, for nearly 20yrs, intel has gone to hardware layer that translates intel instructions into risc micro-ops for execution ... largely negating any risc performance advantage. note that somewhat similar (out-of-order, etc) technology started to be introduced for z196 ... claiming it provided over half the performance improvement from z10 to z196 ... and further additions responsible for some of the z196 to ec12 performance improvement. another technology (compensating for stalled instructions) is hyperthreading. I first ran into it when I was asked to help 370/195 for a hyperthreading implementation they wanted to do. 370/195 had pipeline supporting out-of-order execution that could run at 10mips ... but didn't have branch prediction ... so conditional branches would stall the pipeline ... many codes only ran at 5mips. The idea was to simulate multiprocessor operation with two instruction streams, registers ... but still the same pipeline and execution units (two 5mip instruction steams keeping the 10mip execution units busy). note that it dates back to acs/360 in the late 60s ... see multithreading reference near the end of this article http://people.cs.clemson.edu/~mark/acs_end.html also referenced here http://en.wikipedia.org/wiki/Simultaneous_multithreading SPARC T5 can have 8chips/system, 16cores/chip and 128threads/chip (aka 8threads/core) http://en.wikipedia.org/wiki/SPARC_T5 by comparison, about same time as ec12, e5-2600v1 had two 8core chips for 16cores total and 400-600+ BIPS rating (depending on model) ... compared to max configured (101 processors) EC12 @75BIPS. both e5-2600v1 and ec12 processor chips are done in 32nm technology. intel has a tick-tock chip generation http://en.wikipedia.org/wiki/Intel_Tick-Tock alternates shrinking previous chip design with new technology (tick, e5-2600v2 22nm tech) and then designing new chip for the new technology (tock, e5-2600v3 redesign 22nm). some e5-2600v3 ( v4) discussion http://techgadgetnews.com/2014/09/21/intel-xeon-e5-2600-v3-haswell-ep-workstation-and-server-processors-unleashed-for-high-performance-computing/ E5-2690v1 at 632BIPS, E5-2690v2 at 790BIPS, E5-2690v3 at 996BIPS, E5-2699v3 at 1.321TIPS. http://www.tomshardware.com/reviews/intel-xeon-e5-2600-v3-haswell-ep,3932-7.html note MIPS/BIPS/TIPS are benchmark iterations compared to 370/158 assumed to be 1MIP processor. -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: BDW length vs. Physical Length
cblaic...@syncsort.com (Blaicher, Christopher Y.) writes: ECKD, which is what all modern DASD is, stands for Extended Count Key Data. The 'Extended' refers to the channel commands you can issue, not the devices capabilities. All blocks written to a ECKD device consist of a Count field, an optional Key field and a Data field. The Count field is 8 bytes long and has a format of CCHHRKDD. (Extended format volume count fields are formatted slightly differently, but for basics, this will do.) all modern disk is fixed-block, there hasn't been any real CKD DASD manufactured for decades ... it is all emulated on industry fixed-block disks. ECKD started out when MVS couldn't support FBA ... and they wanted to retrofit 3380 3mbyte/sec disks to 1683033 1.5mbyte/sec channels ... getting CALYPSO to work was something of horror story past posts mentioning FBA, CKD, multi-track seek, etc http://www.garlic.com/~lynn/submain.html#dasd past posts specifically mentioning CALYPSO http://www.garlic.com/~lynn/2004o.html#7 Integer types for 128-bit addressing http://www.garlic.com/~lynn/2007e.html#40 FBA rant http://www.garlic.com/~lynn/2007f.html#0 FBA rant http://www.garlic.com/~lynn/2008q.html#40 TOPS-10 http://www.garlic.com/~lynn/2009k.html#44 Z/VM support for FBA devices was Re: z/OS support of HMC's 3270 emulation? http://www.garlic.com/~lynn/2009p.html#11 Secret Service plans IT reboot http://www.garlic.com/~lynn/2010e.html#36 What was old is new again (water chilled) http://www.garlic.com/~lynn/2010h.html#30 45 years of Mainframe http://www.garlic.com/~lynn/2010n.html#14 Mainframe Slang terms http://www.garlic.com/~lynn/2011e.html#35 junking CKD; was Social Security Confronts IT Obsolescence http://www.garlic.com/~lynn/2012j.html#12 Can anybody give me a clear idea about Cloud Computing in MAINFRAME ? http://www.garlic.com/~lynn/2012o.html#64 Random thoughts: Low power, High performance -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: z/OS MD5 file hashing
002782105f5c-dmarc-requ...@listserv.ua.edu (Frank Swarbrick) writes: Does anyone know of a program/subroutine that can read any kind of MVS sequential dataset and calculate an MD5 hash on it? By any kind I am specifically meaning a file that is either FB or VB and can have any LRECL. note, MD5 has been depreciated for some time (decade ago, I was getting real-time messages from somebody in the crypto rump session where the compromise was being described) ... and asked to do list of internet RFCs mentioning/referencing MD5. http://en.wikipedia.org/wiki/Collision_attack my rfc index http://www.garlic.com/~lynn/rfcietff.htm see special list of RFCs referring to MD5 more recent reference 6151 I Updated Security Considerations for the MD5 Message-Digest and the HMAC-MD5 Algorithms, Chen L., Turner S., 2011/03/06 (7pp) (.txt=14662) (Updates 1321, 2104) (Refs 1321, 1939, 2104, 2202, 4231, 4270, 4493) (Ref'ed By 6150, 6176, 6331, 6421, 6528, 6542, 6668, 6920, 6929, 6931, 6952, 7217, 7292, 7298, 7317, 7321, 7376) from above: 2. Security Considerations MD5 was published in 1992 as an Informational RFC. Since that time, MD5 has been extensively studied and new cryptographic attacks have been discovered. Message digest algorithms are designed to provide collision, pre-image, and second pre-image resistance. In addition, message digest algorithms are used with a shared secret value for message authentication in HMAC, and in this context, some people may find the guidance for key lengths and algorithm strengths in [SP800-57] and [SP800-131] useful. MD5 is no longer acceptable where collision resistance is required such as digital signatures. It is not urgent to stop using MD5 in other ways, such as HMAC-MD5; however, since MD5 must not be used for digital signatures, new protocol designs should not employ HMAC-MD5. Alternatives to HMAC-MD5 include HMAC-SHA256 [HMAC] [HMAC-SHA256] and [AES-CMAC] when AES is more readily available than a hash function. ... snip ... -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Page Data Set Sizes and Volume Types
t...@tombrennansoftware.com (Tom Brennan) writes: Me too - until just a few days ago when I happened upon a number of 3380's defined at a client site. All I can guess is these were still real 3380's at the time they needed to be moved to a DS8000. TASID shows them as 3380-TC3 (whatever that is) at 3,339 cyls. I think I remember a type 3380-K (triple density?), but much of those years is just a blur to me. original 3380 (1981) had twenty track width spacing between (885) tracks ... they then doubled (1770, 3380E, 1985) and then tripled (2655, 3380K, 1987) the number of tracks ... by cutting the inter-track spacing. http://www-03.ibm.com/ibm/history/exhibits/storage/storage_3380.html 3390 announce nov1989 http://www-03.ibm.com/ibm/history/exhibits/storage/storage_3390.html however, as periodically mentioned, there haven't been real CKD DASD manufactured for decades, all just emulation on industry standard fixed-block disks. http://www.garlic.com/~lynn/submain.html#dasd as an aside ... even 3380s CKD were really (32byte) fixed-block http://www.bitsavers.org/pdf/ibm/dasd/reference_summary/GX26-1678-0_3380_Reference_Summary_Feb83.pdf all disk technology was moving to fixed-block by the late 70s ... but MVS inability to come up with fixed-block support required CKD emulation long after CKD was obsolete. there was special 3380j end of 1988 ... which had avg. seek of 12ms and max. seek ms 21 ... compared to 16ms 29ms for 3380k ... but the 3380j had only 885 tracks (same capacity as original 3380) ... one is tempted to believe that the 3380j might have really been a 3380k limited to only accessing 1/3rd of the platter (note seek time isn't strictly linear since there is acceleration latency). recent post mentioning early 80s semi-facetious discussion at SHARE about doing a fast 3380 (with fewer tracks by microcode change). http://www.garlic.com/~lynn/2014m.html#87 Death of spnning disk? From IBM 3380 history reference: In September 1987, IBM announced a significant extension to the 3380 series: the Model K DASD that stored 7.5 billion characters of information, and the densest disk device IBM ever manufactured; and the high-speed Model J, which could locate data faster than any previous 3380 DASD. The Model J found the correct information track in an average time of just 12 thousandths of a second. Customers who installed Model Js, which could store 2.5 billion characters of data, could upgrade it to the denser Model K. ... snip ... -- virtualization experience starting Jan1968, online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN