Is Inter-CP Quiesce time counted as CPU time?
If a z/VM guest partially or completely purges the TLB on a z10 or z196, is the time required to quiesce CPs to coordinate the requested purge counted toward total CPU time for the guest requesting the purge? If so does the guest requesting the purge get tagged for all the CPU time required to coordinate purge operations across all CPs for the z/VM or is the time apportioned by CP to the specific guest active on each CP at the time the purge was requested? If the time isn¹t counted toward CPU time for the guest requesting the purge, how is that allocated? Thanks --. .- .-. -.-- Gary Dennis required for the purge
Seinfeld's Contribution to the The Principles of Operation
There is only one place in z Architecture Principles of Operation where the word substantially is used. Even more surprising than its use is the fact that it appears immediately before the word accurate. My interest in the reference bit, though passing, is sincere since I would like to hang my hat on the validity of that indicator. Substantially means to a great extent or degree. In system z architecture, I am unfamiliar with this level of accuracy. For your reference (no pun intended) the paragraph in z POPS is: The record provided by the reference bit is substantially accurate. The reference bit may be set to one by fetching data or instructions that are neither designated nor used by the program, and, under certain conditions, a reference may be made without the reference bit being set to one. Under certain unusual circumstances, a reference bit may be set to zero by other than explicit program action. Although Seinfeld had to have written the previous 67 words, it is possible (age 21 at the time) for him to have contributed the original paragraph from the 1975 POPS and simply updated the paragraph for System z. The 1975 paragraph follows. The record of references provided by the reference bit is substantially accurate. The reference bit may be turned on by fetching data or instructions that are neither designated nor used by the program, and, under certain conditions, a reference may be made without the reference bit being turned on. Under certain unusual conditions, a reference bit that is on may be turned off by other than explicit program action. Turned on and Turned off apparently gave way to the more precise set to one and set to zero for those so technically challenged that on/off had lost all meaning. Think about it; the technical writers were striving for more precision in their description than the thing being described actually provided. Does anyone have a clue what sentences 2 and 3 (substantially) mean in either paragraph and would anyone venture a guess as to why, after 40 years of architectural excellence, the reference bit can't be more than substantially accurate? My curiosity is killing me. --. .- .-. -.-- Gary Dennis
Maximum Virtual Storage
What is the maximum guest virtual storage supported by z/VM? --. .- .-. -.-- Gary Dennis Mantissa Corporation
Re: Maximum Virtual Storage
The thread on mixed paging volumes caused me to ask the question. I should have been more specific. If the volume limit for a z/VM page volumes is 240+, how does this relate to maximum defined virtual storage for all active guests under a z/VM image? For example, in an environment where each guest requires 4BG virtual, how many such guests could a single z/VM system manage? On 10/1/10 3:07 PM, Mike Walter mike.wal...@hewitt.com wrote: Do you mean REAL virtual storage, to which answers have already been supplied? Or do you mean VIRTUAL virtual storage, as documented as the Maximum Input Values for Storage Units in the CP Planning and Administration manual? For z/VM 5.4 and 6.1 the maximum stor size for any virtual machine is 16E (exabytes). I'd venture a guess that IBM would be pleased to sell you sufficient real storage and DASD to support a few of those VMs, their paging and dump space requirements. :-) 11010100100010011001001011010100111001101001100100111010001111 0110011001 Mike Walter Hewitt Associates The opinions expressed herein are mine alone, not my employer's. Gary M. Dennis gary.den...@mantissa.com Sent by: The IBM z/VM Operating System IBMVM@LISTSERV.UARK.EDU 10/01/2010 02:47 PM Please respond to The IBM z/VM Operating System IBMVM@LISTSERV.UARK.EDU To IBMVM@LISTSERV.UARK.EDU cc Subject Maximum Virtual Storage What is the maximum guest virtual storage supported by z/VM? --. .- .-. -.-- Gary Dennis Mantissa Corporation The information contained in this e-mail and any accompanying documents may contain information that is confidential or otherwise protected from disclosure. If you are not the intended recipient of this message, or if this message has been addressed to you in error, please immediately alert the sender by reply e-mail and then delete this message, including any attachments. Any dissemination, distribution or other use of the contents of this message by anyone other than the intended recipient is strictly prohibited. All messages sent to and from this e-mail address may be monitored as permitted by applicable law and regulations to ensure compliance with our internal policies and to protect our business. E-mails are not secure and cannot be guaranteed to be error free as they can be intercepted, amended, lost or destroyed, or contain viruses. You are deemed to have accepted these risks if you communicate with us by e-mail. --. .- .-. -.-- Gary Dennis Mantissa Corporation 1121 Edenton Street Birmingham, Alabama 35242-9257 0 ... living between the zeros... 0 p: 205.968-3942 m: 205.218-3937 f: 205.968.3932 gary.den...@mantissa.com http://www.mantissa.com http://www.idovos.com
Re: Mixed page volume sizes
S if you had guests averaging 18GB each, and you follow recommendations for page volume utilization (50% as I understand it), Mod 9s would yield around 4GB useable page space each. That would give you 100 such images per VM .. (250 volumes times the 4GB/volume divided by 18GB per image)? I know this rough but am I headed in the right direction? I'm making the assumption (an this may be incorrect) that the available real storage backing could support virtual requirements. On 9/30/10 9:24 AM, Rob van der Heij rvdh...@gmail.com wrote: On Thu, Sep 30, 2010 at 3:39 PM, George Henke/NYLIC george_he...@newyorklife.com wrote: Sorry for my ignorance and naivete, but I am simply staggered at the volume of page volumes of whatever size, 100 - 240. I have just a Dev Test z/VM environment with 2 measly 3390-3's for Paging. Granted, I am running only 5 z/OS vm's and 3 Linux vm's. My test system does not have that either. But think of 250G real memory and some 100 Linux guests of 4-10 GB each, all running enterprise applications. | Rob --. .- .-. -.-- Gary Dennis Mantissa Corporation 1121 Edenton Street Birmingham, Alabama 35242-9257 0 ... living between the zeros... 0 p: 205.968-3942 m: 205.218-3937 f: 205.968.3932 gary.den...@mantissa.com http://www.mantissa.com http://www.idovos.com
Re: ACM award - they deserve it....
Less than an hour before reading this thread I had IPLed two x86 operating systems under z/VM 5.3. One is a tiny OS used as a checkpoint after major changes to the x86 virtualization layer. It's just a way to make sure the puzzle is still assembled more or less correctly. The other OS is DSL. The IPL time for DSL up to the line-mode prompt is now 5 seconds with zero tuning in any of the z86VM components. Using VNC the GUI for each OS is live-scroll capable over a VPN connection to the IBM development site in Dallas. The current x86 virtualization marketplace may not be as predictable (even for bean counters) as one might imagine. Those same bean counters will be the first ones to ask Why not 1 box instead of 50? just like they asked Why not 50 boxes instead of 500?. Wouldn't that be preferable to an ACM award? --. .- .-. -.-- Gary Dennis Mantissa Corporation 0 ... living between the zeros... 0 On 3/31/10 10:32 AM, Dave Wade g4...@dpwade.eclipse.co.uk wrote: In my humble opinion the main reason VMWare (an to a lesser extent HyperV) is popular at present is because it allows bean counters to demonstrate huge instant savings. Where I work we have around 200 Windows servers, many were bought around 5 years ago so will need replacing soone. In general we have a separate server not for performance reasons but more for separation of control and software options. Based on a limited trial I would say we could consolidate 75% of these servers at a rate of at least 10 to 1 using VMWare, and still have enough headroom to loose a physial server with no performance impact. So that's take the 150 lowest loaded servers and replace them with 15 servers running VMWare. To a bean counter that's a 90% reduction in power consumption, a 90% reduction in floor space, and a 90% reduction in hardware support costs.I am sure some think that should also be a 90% reduction in support staff, but of course that's not true. Whilst VMWare is fun to manage, it needs managing and also capacity planning. In practice the reduction is some what less than 90%. . To use the vernacular, a VMWare server will be a fully loaded server with multiple CPU's, lots of RAM, multiple SAN and Network interfaces for load balancing and resilience. In order to fit these in it will be a 2U server and some of our existing are 1U, on the other hand others are 4U... BUT there will be a big saving. Now compare that with zVM. With that you were frugal from day1 so there aren't any savings. So the bean counters can't show cost reductions, so they don't like it utterly blinkered Dave. - Original Message - From: Barton Robinson bar...@vm1.velocity-software.com To: IBMVM@LISTSERV.UARK.EDU Sent: Wednesday, March 31, 2010 3:53 PM Subject: Re: ACM award - they deserve it If you go to conferences such as CMG (Computer Management Group), that has been a mainframe organization (meaning MVS or z/OS) since it started, our VM has never been represented, but VMWare now has many sessions. It's depressing to see 80 people in entry level performance session for VMWare and no z/VM sessions on the agenda of a mainframe conference. Early this year I was hearing ads for VMWare on the local radio station. I can only assume that VM is being outmarketed worldwide (or at least that VMWare is being marketed worldwide and VM is not marketed publicly at all). It doesn't matter if our mousetrap is better if nobody is out there trying to get mindshare (marketing). Preaching/grumbling to the choir doesn't change anything. So when was the last time that any of you tried to get a case study published showing how great your accomplishments are using z/VM? There are very few published stories (sorry games on z don't impress bean counters or executives, it's rather demeaning), we need REAL business case studies showing the value of z/VM to real companies. If we get enough and executives do a google search on VM, maybe they will find something useful? Bill Munson wrote: Jim, You are right, that makes me mad also. IBM really blew it when they did not trade mark VM munson Jim Elliott jelli...@gdlvm7.vnet.ibm.com Sent by: The IBM z/VM Operating System IBMVM@LISTSERV.UARK.EDU 03/30/2010 09:34 PM Please respond to The IBM z/VM Operating System IBMVM@LISTSERV.UARK.EDU To IBMVM@LISTSERV.UARK.EDU cc Subject Re: ACM award Today the Association for Computing Machinery (of which I have been a member since 1970) made the following award: VMware Workstation 1.0, the Software System Award, for bringing virtualization technology to modern computing environments, spurring a shift to virtual-machine architectures, and allowing users to efficiently run multiple operating systems on their desktops. Aside from the run multiple OSes on the desktop part, shouldn't we be insulted? Chip: Yes, we should be insulted. I remember being very upset the
VPN into Second Level VM
We are developing on a second level VM system with access through VPN. Regardless of the VPN client location (Home, office ), ISP download/upload speed and chip speed of the client machine, we can¹t seem to get any more than 135K bytes per second over the VPN. This number holds true for both FTP operations and VNC client connections to second level VM guests. Any thoughts or suggestions what we might check or tweak in the second level VM system to improve this speed.. Thanks --. .- .-. -.-- Gary Dennis Mantissa Corporation
Fixed length field alignment degradation
The POPS For System z contains the following note: Programming Note: For fixed-field-length operations with field lengths that are a power of 2, significant performance degradation is possible when storage operands are not positioned at addresses that are integral multiples of the operand length. To improve performance, frequently used storage oper- ands should be aligned on integral boundaries. Does anyone know what significant performance degradation means in terms of true machine overhead? Is anyone aware of published degradation numbers for non-aligned fixed length field operations? --. .- .-. -.-- Gary Dennis Mantissa Corporation 0 ... living between the zeros... 0
Re: VM Best Practices
In the movie In Search of the Holy Grail, the knights who say nee could not bear to hear a certain words. One of those words was it. It appears that best practice may be not to ask about - (words redacted). Thanks to all who responded. --. .- .-. -.-- Gary Dennis Mantissa Corporation 0 ... living between the zeros... 0
VM Best Practices
Watching ³A Christmas Story² makes me wonder if you can ³shoot your eye out² through errors of omission with a VM system. Can anyone point me to a source for z/VM ³Best Practices² that addresses low level system recovery (essentially disaster recovery). Thanks Gary Dennis Mantissa Corporation 0 ... living between the zeros... 0
Re: Questions - zVM Limits/Hardware Support
Extremely useful. The limits document by by Bill Bitner and another by the same author on performance are great resources. If you attempted to configure a VM system with 700 active users each having an average working set of 8GB each what kind of train wreck could that create? All recommendations I see call for keeping the virtual to real ratio 3:1 and I wonder what would happen in this case. I note that VM has been tested to the 1TB real level. If the max number of paging volumes is 255 but you should keep the utilization for of each device at 50% for optimal performance, what does that mean? Are we talking 50% of the potential sustained I/O rate for the device or 50% of physical capacity. If the latter, that pushes the paging cap down to 5.6/7.95. What transmission limits are imposed by ISFC? When would the CTC link max out? Could a very high activity IUCV communications link function well over ISFC? --. .- .-. -.-- Gary Dennis Mantissa Corporation 0 ... living between the zeros... 0 On 10/28/09 4:06 PM, Alan Altmark alan_altm...@us.ibm.com wrote: On Wednesday, 10/28/2009 at 04:28 EDT, Gar over ISFC? M. Dennis gary.den...@mantissa.com wrote: What is the maximum page space supported by zVM? What is the supported real storage limit for the hardware and for VM itself? Max page space: - For ECKD: 11.2 TB - For FCP: 15.9 TB (emulated FBA on SCSI) Note that optimal performance requires that you keep utilization of each device to 50%. Memory limits: - z/VM supports an LPAR up to 256 GB in size - The amount of memory on the box and in an LPAR depends on the hw - Biggest z10 has up to 1.5 TB memory and a the largest LPAR can be 1.0 TB You can find these and other z/VM limits in Bill Bitner's z/VM Limits presentation at http://www.vm.ibm.com/devpages/bitner/presentations/vmlimits.pdf What is the highest total sustained I/O rate you have witnessed on a VM system? I have heard rumors of a z/VM paging rate of 200K pages/second on a robust I/O subsystem, but I don't know if that's what you're referring to. Guest I/O data rates are a function of access to the CPU and the size of the I/O operation, so it depends. I see from an BM presentation that support is available for simulated guest coupling. Does zVM support a real coupler facility for intersystem (VM SYSTEMS) communications? z/VM does not use and does not allow guest access to the real Coupling Facility. z/VM's native intersystem comms mechanism is ISFC, based on CTCs. Alan Altmark z/VM Development IBM Endicott --. .- .-. -.-- Gary Dennis Mantissa Corporation 0 ... living between the zeros... 0
Questions - zVM Limits/Hardware Support
Questions... What is the maximum page space supported by zVM? What is the supported real storage limit for the hardware and for VM itself? Does anyone have experience with a zVM system having over 500 non-CMS guest systems? What is the highest total sustained I/O rate you have witnessed on a VM system? I see from an BM presentation that support is available for simulated guest coupling. Does zVM support a real coupler facility for intersystem (VM SYSTEMS) communications? Thanks --. .- .-. -.-- Gary Dennis Mantissa Corporation
Chunk (RAW Block) access to DS8000 storage
Does anyone know of a way to access very large chunks of DS8000 storage without minidisk definitions? We are looking for a way to allocate substantial increments of storage (several terabytes at a time) with the least administrative overhead. --. .- .-. -.-- Gary Dennis Mantissa Corporation
Re: Shared File System Interface
³Twitchy² - is a squirrel and best friend of Wolf W. Wolf in Hoodwinked. ³twitchy² as used below makes me, well Twitchy. Could you elaborate on why use of SFS via NFS ³Sort of makes the z guys twitchy² --. .- .-. -.-- Gary Dennis Mantissa Corporation On 5/1/09 7:39 AM, Dean, David (I/S) david_d...@bcbst.com wrote: Yep. Then you can Samba out the NFS to window's boxes and have a really nice file server. Sort of makes the z guys twitchy though. David Dean Information Systems *bcbstauthorized* -Original Message- From: The IBM z/VM Operating System [mailto:ib...@listserv.uark.edu] On Behalf Of Alan Altmark Sent: Wednesday, April 29, 2009 1:15 PM To: IBMVM@LISTSERV.UARK.EDU Subject: Re: Shared File System Interface On Wednesday, 04/29/2009 at 01:06 EDT, Dave Jones d...@vsoft-software.com wrote: Nope, afraid notbut it would be way cool if Linux, as a guest of z/VM, could read/write SFS directories and files. It can. It just needs to use NFS to do it. Alan Altmark z/VM Development IBM Endicott - Please see the following link for the BlueCross BlueShield of Tennessee E-mail disclaimer: http://www.bcbst.com/email_disclaimer.shtm
Shared File System Interface
Is there any documented APPC interface to SFS for non-CMS operating systems? --. .- .-. -.-- Gary Dennis
Re: *SIGNAL Service Experience
Only A to B as we have no broadcast requirement. --. .- .-. -.-- Gary Dennis Mantissa Corporation On 4/21/09 4:08 PM, Alan Altmark alan_altm...@us.ibm.com wrote: On Tuesday, 04/21/2009 at 01:55 EDT, Gary M. Dennis gary.den...@mantissa.com wrote: The IUCV tests seemed insensitive to transmission size. Whether we sent 8 bytes or 32,000, the 5 to 7 second window held. The savings of *SIGNAL would be the use of the broadcast function. Instead of UserA sending to UserB, then UserC, then UserD, and so on, userA just broadcasts the IUCV data, saving an IUCV SEND operation as well as the logic to keep track of B, C, and D (which *SIGNAL can help with, too). For two guests, the savings would be zero. What scenario did you compare? Alan Altmark z/VM Development IBM Endicott
2A8 Diagnose Undocumented return code
While attempting a ³Establish Device Connection² we receive a CC 2 with a return code of x¹10¹ Anyone know what this means? The documentation is the most recent CP programming services. Thanks --. .- .-. -.-- Gary Dennis Mantissa Corporation
Re: *SIGNAL Service Experience
Thanks. The following is from some tests we conducted to compare *SIGNAL to straight IUCV. 2097 Dallas Development, second level VM, under CMS. 100,000 signal / acknowledgement transmissions. The straight IUCV test sent only 8 bytes to try and put it on equal footing with *SIGNAL. 4 *SIGNAL tests - ran 5 seconds each over a 2 minute period. 4 IUCV tests - ran 5, 7, 7, 6, 7 seconds over a 2 minute period. The IUCV tests seemed insensitive to transmission size. Whether we sent 8 bytes or 32,000, the 5 to 7 second window held. --. .- .-. -.-- Gary Dennis Mantissa Corporation On 4/20/09 12:53 AM, Alan Altmark alan_altm...@us.ibm.com wrote: On Friday, 04/17/2009 at 12:16 EDT, Gary M. Dennis gary.den...@mantissa.com wrote: IF you have experience using *SIGNAL service for high volumes To my knowledge, the only exploiter of *SIGNAL is GCS, which has been used since around 1986 to run VTAM, NetView, RSCS, AVS, PSF and VSCS applications. The signalling GCS performs is primarily used to - Signal the presence of data in shared memory (usually so that the target of the signal can send it.) - Run a named entry point in another group member (usually so target of the signal can receive data placed in shared memory) - Dump shared memory if a member of the group dies - Join or remove a member of the group GCS has handled large workloads for decades without incident. Alan Altmark z/VM Development IBM Endicott
*SIGNAL Service Experience
IF you have experience using *SIGNAL service for high volumes THEN Could you provide information (or simply observation) relating to overhead, latency? ELSEIF You know compelling reasons this service should not be considered for very high signal volumes THEN This is an excellent opportunity to keep fellow primates out of harms way. ENDIF Thanks --. .- .-. -.-- Gary Dennis Mantissa 0 ... living between the zeroes... 0
Re: File System - If you had everything, where would you put it?
Jeff, What I was trying to determine if there was a way to use ZFS on OpenSolaris System z as a high speed space management vehicle while bypassing conventional transport layers? For example, let¹s say there existed a way to push data to the IO appliance cross-memory (guest to ZFZ, ZFS to guest) such that an interface in the appliance could act as a local proxy for each guest using the service. Regards, Gary --. .- .-. -.-- 0 ... living between the zeroes... 0 On 4/3/09 11:29 AM, Jeff Savit jsa...@gmail.com wrote: Gary M. Dennis wrote: 1. How does OpenSolaris zfs utilize the storage tier on System z? Are the disks allocated to zfs pool(s) simply reserved CMS formatted disks? Exactly that: minidisks used by OpenSolaris on z port (sirius for short) are CMS-formatted and RESERVE-d minidisks. That way sirius knows which minidisks to *not* touch. 2. How does the Async I/O in ZFS work? Would the guest that requested the I/O be signaled with an ext interrupt by the I/O appliance? Do you mean from the POSIX API level or at the virtual machine I/O level? ZFS obeys standard POSIX semantics for file I/O, including the ability to specify sync semantics. From the physical I/O perspective, everything is done by DIAG 250, which permits non-guest-blocking I/O. 3. What API/system transport layer would be used in VM for guests to conduct I/O through an appliance managing ZFS? Not sure what you mean by this - could you explain? Guests running which OS talking to an appliance running where? regards, Jeff
Diagnose X'248' - Copy to Primary
Can anyone comment on the overhead (or lack thereof) relating to use of this diagnose. Thanks --. .- .-. -.-- Gary Dennis 0 ... living between the zeroes... 0
Re: File System - If you had everything, where would you put it?
Either would work, but a shared address space would seem to be the most lightweight approach. On 4/3/09 3:57 PM, Dave Jones d...@vsoft-software.com wrote: Would you consider a mechanism like IUCV or shared data address spaces as acceptable methods to send data back and forth between z/VOS guests and a file server of some sort? Gary M. Dennis wrote: Jeff, What I was trying to determine if there was a way to use ZFS on OpenSolaris System z as a high speed space management vehicle while bypassing conventional transport layers? For example, let¹s say there existed a way to push data to the IO appliance cross-memory (guest to ZFZ, ZFS to guest) such that an interface in the appliance could act as a local proxy for each guest using the service. Regards, Gary --. .- .-. -.-- 0 ... living between the zeroes... 0 On 4/3/09 11:29 AM, Jeff Savit jsa...@gmail.com wrote: Gary M. Dennis wrote: 1. How does OpenSolaris zfs utilize the storage tier on System z? Are the disks allocated to zfs pool(s) simply reserved CMS formatted disks? Exactly that: minidisks used by OpenSolaris on z port (sirius for short) are CMS-formatted and RESERVE-d minidisks. That way sirius knows which minidisks to *not* touch. 2. How does the Async I/O in ZFS work? Would the guest that requested the I/O be signaled with an ext interrupt by the I/O appliance? Do you mean from the POSIX API level or at the virtual machine I/O level? ZFS obeys standard POSIX semantics for file I/O, including the ability to specify sync semantics. From the physical I/O perspective, everything is done by DIAG 250, which permits non-guest-blocking I/O. 3. What API/system transport layer would be used in VM for guests to conduct I/O through an appliance managing ZFS? Not sure what you mean by this - could you explain? Guests running which OS talking to an appliance running where? regards, Jeff --. .- .-. -.-- Gary Dennis Mantissa Corporation 1121 Edenton Street Birmingham, Alabama 35242-9257 0 ... living between the zeroes... 0 p: 205.968-3942 m: 205.218-3937 f: 205.968.3932 gary.den...@mantissa.com http://www.mantissa.com http://www.idovos.com
Re: File System - If you had everything, where would you put it?
Something along these lines Guests pull on read Servers pull on write Async only --. .- .-. -.-- Gary Dennis Mantissa 0 ... living between the zeroes ... 0 On 4/3/09 4:26 PM, Alan Altmark alan_altm...@us.ibm.com wrote: On Friday, 04/03/2009 at 03:38 EDT, Gary M. Dennis gary.den...@mantissa.com wrote: What I was trying to determine if there was a way to use ZFS on OpenSolaris System z as a high speed space management vehicle while bypassing conventional transport layers? For example, let?s say there existed a way to push data to the IO appliance cross-memory (guest to ZFZ, ZFS to guest) such that an interface in the appliance could act as a local proxy for each guest using the service. Just in case you are trying to connect Diag 0x248 with push data to OpenSolaris, don't bother. First, Diag 0x248 is a read-only function. Second, OpenSolaris, like Linux, is DAT ON. That means they cannot create or access Data Spaces. Your only shared-memory solution is a DCSS. Alan Altmark z/VM Development IBM Endicott
Re: File System - If you had everything, where would you put it?
As it is written. Guests pull from the server on read requests and servers pull from the guests on write requests. We seem to be missing an interrupt sequence don¹t we? Gary On 4/3/09 6:15 PM, Jeff Savit jsa...@gmail.com wrote: Do you mean 'pull' or 'poll' on read, or 'push' on write? :-) In any case, Alan is right, and the lowest latency way for virtual machin es to share data is with a DCSS. cheers, Jeff On Fri, 3 Apr 2009 16:47:58 -0500, Gary M. Dennis gary.den...@mantissa.c om wrote: Something along these lines Guests pull on read Servers pull on write Async only --. .- .-. -.-- Gary Dennis Mantissa 0 ... living between the zeroes ... 0 On 4/3/09 4:26 PM, Alan Altmark alan_altm...@us.ibm.com wrote: On Friday, 04/03/2009 at 03:38 EDT, Gary M. Dennis gary.den...@mantissa.com wrote: What I was trying to determine if there was a way to use ZFS on OpenSolaris System z as a high speed space management vehicle while bypassing conventional transport layers? For example, let?s say there existed a way to push data to the IO appliance cross-memory (guest to ZFZ, ZFS to guest) such that an interface in the appliance could act as a local proxy for each guest using the service. Just in case you are trying to connect Diag 0x248 with push data to OpenSolaris, don't bother. First, Diag 0x248 is a read-only function. Second, OpenSolaris, like Linux, is DAT ON. That means they cannot cr eate or access Data Spaces. Your only shared-memory solution is a DCSS. Alan Altmark z/VM Development IBM Endicott =
Re: File System - If you had everything, where would you put it?
David, Thanks. 1. How does OpenSolaris zfs utilize the storage tier on System z? Are the disks allocated to zfs pool(s) simply reserved CMS formatted disks? 2. How does the Async I/O in ZFS work? Would the guest that requested the I/O be signaled with an ext interrupt by the I/O appliance? 3. What API/system transport layer would be used in VM for guests to conduct I/O through an appliance managing ZFS? Regards, --. .- .-. -.-- Gary Dennis 0 ... living between the zeroes ... 0 On 4/1/09 8:47 PM, David Boyes dbo...@sinenomine.net wrote: Some of the STK/Sun disks have hardware features to do this. OpenAFS or Lustre could do this if you allow Linux guests to provide the services, or ZFS if you allow OpenSolaris guests. It¹d be very easy to package up an appliance server image to do what you need done with either one (Linux or OpenSolaris). Other than that, you have to use SFS or BFS, and both need some work to do continuous availability. Minidisks won¹t work for this. On 4/1/09 5:32 PM, Gary M. Dennis gary.den...@mantissa.com wrote: Is there a VM I/O management system available which will: 1. Support space allocation requests for guests on a sparse basis? The file server needs to make the guest believe it actually has the entire allocation while only tying up space in the pool the guest actually used. 2. Support Async I/O requests from multiple guests? 3. Dynamically scale and from sub- TB to n EB.
File System - If you had everything, where would you put it?
Is there a VM I/O management system available which will: 1. Support space allocation requests for guests on a sparse basis? The file server needs to make the guest believe it actually has the entire allocation while only tying up space in the pool the guest actually used. 2. Support Async I/O requests from multiple guests? 3. Dynamically scale and from sub- TB to n EB. --. .- .-. -.-- Gary Dennis 0 ... living between the zeroes ... 0
Diagnose x'250' / z/VM I/O scheduling
Given A file pool consisting of 5 VDEV files on 5 separate real devices 2 cylinders per device 4096 block size When: a block chain is given Diag x'250' (async) for either read or write such that 4 blocks are written to or read from each track within the 5 files. Question(s): Does the 250 interface make any attempt to optimize I/O operations by constructing chained channel programs for single-track or consecutive-track multi-record writes/reads? If that is not the case, is such optimization achieved at a more basic level in z/VM real device I/O scheduling? Curiosity killed the In either of the above cases (that is if channel programs are chained based on intra-request I/O patterns), will either 250 or VM perform inter-request channel program chaining for multiple async requests targeting the same real device? Thanks --. .- .-. -.-- Gary Dennis 0 ...living between the zeroes... 0
Re: Diagnose x'250' / z/VM I/O scheduling
Thank you. Let me clarify. The questions aren't about the device which, as you correctly observe, is now a software figment. All we see and can control on the System z end is the sub-channel. My thinking (and this may be entirely off base) is that scheduling a sub-channel request to write 25 blocks via a composite channel program is somehow better than 25 separate requests. Regards, Gary On 3/30/09 10:22 AM, Schuh, Richard rsc...@visa.com wrote: The real question is, with today's disk arrays, what really is the optimal order of the CCWs in the chain? What may seem logical to you, who have apparently been around long enough to remember the days of each disk being a physical unit with cylinders and tracks being arranged sequentially, may not be optimal for disks that are striped across many physical disks. The VSSI products, VPARS and VTAPE, use the BLOCKIO routines and, if I am not mistaken, the VSSI code optimizes the channel programs by sorting them into sequential order before the DIAG is issued. Regards, Richard Schuh -Original Message- From: The IBM z/VM Operating System [mailto:ib...@listserv.uark.edu] On Behalf Of Gary M. Dennis Sent: Monday, March 30, 2009 7:39 AM To: IBMVM@LISTSERV.UARK.EDU Subject: Diagnose x'250' / z/VM I/O scheduling Given A file pool consisting of 5 VDEV files on 5 separate real devices 2 cylinders per device 4096 block size When: a block chain is given Diag x'250' (async) for either read or write such that 4 blocks are written to or read from each track within the 5 files. Question(s): Does the 250 interface make any attempt to optimize I/O operations by constructing chained channel programs for single-track or consecutive-track multi-record writes/reads? If that is not the case, is such optimization achieved at a more basic level in z/VM real device I/O scheduling? Curiosity killed the In either of the above cases (that is if channel programs are chained based on intra-request I/O patterns), will either 250 or VM perform inter-request channel program chaining for multiple async requests targeting the same real device? Thanks --. .- .-. -.-- Gary Dennis 0 ...living between the zeroes... 0 --. .- .-. -.-- Gary Dennis Mantissa Corporation 1121 Edenton Street Birmingham, Alabama 35242-9257 0 ... living between the zeroes... 0 p: 205.968-3942 m: 205.218-3937 f: 205.968.3932 gary.den...@mantissa.com http://www.mantissa.com http://www.idovos.com
Thanks All
On behalf of the zVos development team I want to express our thanks to the membership of this list. Many of your discussions have helped our effort enormously. When I joined this list almost exactly a year ago, I did so with the anticipation that we might need help if we got behind the curve technically. I missed the mark (and the point altogether) though that narrowly defined rationale for joining the list was valid. The knowledge reservoir and been-there-done-that experience available in this forum is remarkable and much appreciated. --. .- .-. -.-- Gary Dennis Mantissa Corporation
CMS threading, Preemption, Involuntary vs Voluntary loss of control
CMS Multitasking Application Programming documentation draws a distinction between voluntary and involuntary loss of control by a thread. Blocking and yielding in a thread are easily understood as causes for voluntary loss of control. Preemption would appear to be just as easily understood. However, the observed behavior of a running application doesn¹t appear to obey the documentation which states in part that at the CMS thread level you can involuntarily (preemptively) lose control to threads in different classes. Alternatively stated, a compute bound thread in the same class as three siblings will never cede control to one of the three siblings but should cede control to threads of higher priority outside it dispatch class. Thread priorities can be established at thread creation time as a plus or minus offset of the creating thread priority. I have an application which creates 6 threads all all of which are given a distinct classes and various priorities at the point of creation. The lowest priority thread is the most compute intensive. All threads other than the compute bound thread exhibit blocking or yielding behavior necessary for voluntary loss of control; the compute bound thread does not. The compute bound thread is the only one that gets dispatched. Preemption at the virtual machine level works beautifully or this would be a really small forum. Given the content of the manual I would expect that the compute bound thread would involuntarily lose control at some point to the higher priority threads in different classes. Questions: 1. Does CMS preemptively dispatch threads as the documentation implies? 2. If so, is there some VM or CMS configuration option that is used to ensure this happens? Thanks --. .- .-. -.-- Gary Dennis Mantissa Corporation
.zVos System z/VM Connectivity - Men or Mice
Over the last few weeks we have determined (without equivocation) that we are z/VM communications mice. Communications men or women would have dispatched the following problem forthwith. What we are attempting to do is describe our connectivity requirements for testing zVos to IBM¹s Dallas development center. I feel as though we are throwing them somewhat of a curve ball because our environment is not Linux and not strictly CMS/VM. Here is what we have: 1. A second level VM system running on the Dallas First level. 2. On that second level system there are two server machines which have to be able to communicate with a z/Vos management console in Birmingham via HTTP. 3. On the same machine there are 3 virtual x86 machines that must be reached from the Birmingham network with RDP and/or VNC over a Specific subnet and port range. 4. The servers and virtualized machines are on the same VLAN. 5. VPN Questions: 1. What has to happen on the first level machine to permit access to a specific subnet and port range on the VLAN of the second level machine? 2. Shouldn't we be able to get away with just having a VSWITCH/VLAN combination on the second level machine without dedicating an OSA adapter? Anybody done this in a DR environment. 3. Would use of a virtual OSA adapter in the second level machine make this easier or just ratchet the level of complexity? Thanks --. .- .-. -.-- Gary Dennis Mantissa Corporation
Re: Value added by z/VM versus VMWARE
If z/VM supported virtual x86 systems, that support would make the platforms extremely competitive and, potentially, cause a sea change in the source of computing resource for x86. Considering the average CPU utilization for x86 desktop systems (less than 15% by some estimates), such support could make for a good match; guest systems that do practically nothing and a virtualization system with a remarkable ability to allocate resources among a large number of guests. On 11/2/08 2:12 PM, Paul Raulerson [EMAIL PROTECTED] wrote: That's what confuses me- the two platforms, mainframe and x86 are hardly competitive to each other. --. .- .-. -.-- Gary Dennis
Re: Virtual machine 'escape'
What effect would this same hack have on the intended target if the x86 system being targeted was running as a guest under z/VM? Wouldn't the ill effects be reduced by the wall between virtual guests inherent with z/VM? On 11/4/08 11:42 AM, David Boyes [EMAIL PROTECTED] wrote: It seems our colleagues doing virtualization on Intel have another possible security concern to worry about now. By far the biggest concern related to virtual machine security is the threat of a virtual machine escape. A virtual machine escape is a theoretical type of attack in which an attacker uses a vulnerability within a virtual machine to take control of either the underlying host operating system, or the hypervisor itself. Upon doing so, the attacker could potentially gain control of the other virtual machines hosted on the server. Why is it such a threat? It's the fear of the unknown, that eventually someone will be able to do it. Not just possible; proven. It's been done on an Intel Pacifica chipset, and there was an excellent paper in IEEE Transactions on Computer Systems on how it was done. --. .- .-. -.-- Gary Dennis Mantissa Corporation 1121 Edenton Street Birmingham, Alabama 35242-9257 p: 205.968-3942 m: 205.218-3937 f: 205.968.3932 [EMAIL PROTECTED] http://www.mantissa.com http://www.idovos.com
Re: VM Virtual CPUs and Threaded CMS Applications
I appreciate your insight. When you state: ³ If you¹re not invoking CMS services from non-base threads² What precisely do you mean by CMS services? Are you referring specifically to the services defined in ³CP Programming Services² OR any call to CMS? This has the potential to derail what we are trying to achieve with CMS so I want to be absolutely sure I understand what you mean. --. .- .-. -.-- Gary Dennis On 10/19/08 2:35 PM, Gillis, Mark [EMAIL PROTECTED] wrote: I haven¹t experienced this specific problem because IBM strongly advised us to not allocate more than 1 virtual CPU to a mutitasking CMS application. The reason they gave was that any CMS services called from a thread running on a non-base CPU would need to be scheduled to run on the base CPU, so that the overheads of this would outweigh the benefits of the extra processors. If you¹re not invoking CMS services from non-base threads then I guess that this won¹t be an issue for you. Mark Gillis Principal Software Engineer Tel: +61 2 8898 2678 Fax: +61 2 8898 2600 [EMAIL PROTECTED] From: The IBM z/VM Operating System [mailto:[EMAIL PROTECTED] On Behalf Of Gary M. Dennis Sent: Sunday, 19 October 2008 10:24 AM To: IBMVM@LISTSERV.UARK.EDU Subject: VM Virtual CPUs and Threaded CMS Applications If you do not have experience with threaded CMS application development, I suggest you read anything but the balance of of this email. I have an application that runs under CMS and consists of three distinct layers. 1. The top layer is some virtualized x86 OS. 2. The middle layer performs x86 to z translation 3. The base layer is everything else. That includes code fragment storage, aging, retrieval, statistics collection/ push using IUCV, etc. Layer two has been developed in such a way that, without layer three, it simply translates a code fragments to z architecture code, executes that code, then discards the translated fragment. It detects the interface stub for layer three and, if that is present, it takes advantage of the capabilities including prior translation reuse. Layer 3 is multithreaded and is the cause/source of the problem. Whether layer 3 is run with layers 1 and 2 or in standalone test mode the results are the same. First the environment: VM 4.3 Number of processors: 2 Virtual CPUs (from 2 to 6 .. See note below) Now the application from 10,000 feet: Layer three consists of a parent thread that creates 4 additional threads. Each thread is created in a dispatch class that is unique. Routines are not shared between threads. Upon entry into each routine, the preamble is destroyed and restored on exit to trap any potential inadvertent share. Critical fields shared between threads are protected by a compare and swap spin lock. Part of the testing consists of pushing 1WAY IUCV messages from each connected client every 20 milliseconds. The VM directory for each of 4 machines (one server and three clients) defines the machine as an XC mode machine with: CPU 00 BASE CPU 01 As each thread is created it requests either BASE or ANY CPU affinity. BASE affinity is reserved for the parent and IUCV message handler threads . ANY is used for all other threads . Each affinity request receives a normal return code. All this works beautifully for days and millions of messages UNTIL the number of virtual CPUs defined exceed the number of real CPUs assigned to the VM image. When this takes place, everything comes unstuck. By everything I mean everything in CMS. Stack overflow (03FF abend), free storage management failure, all of it. The multitasking application dev guide states that to the extent possible, dispatch classes are assigned to vCPUs and further states that the max number of vCPUs that may be utilized is equal to the number of dispatch classes. Whether the vCPUs are defined in the user directory entry OR they are created dynamically using the CPU Create CMS function, the results are the same. My questions) 1. Has anyone had a similar experience? 2. Is this a known issue with 4.3? Or in more current releases? 3. Although this seems to be telling me no to go there, I¹ve tried but cannot find anything that says ³You¹ll shoot your eye out , kid.² if you define more virtual CPUs than real processors. Anyone know of such a restriction? 4. Is it possible that CMS kernel services don¹t tolerate a situation where the number of virtual CPUs exceeds ³real² processors? Thanks in advance for any insight you might have on this behavior. --. .- .-. -.-- Gary Dennis Mantissa Corporation
VM Virtual CPUs and Threaded CMS Applications
If you do not have experience with threaded CMS application development, I suggest you read anything but the balance of of this email. I have an application that runs under CMS and consists of three distinct layers. 1. The top layer is some virtualized x86 OS. 2. The middle layer performs x86 to z translation 3. The base layer is everything else. That includes code fragment storage, aging, retrieval, statistics collection/ push using IUCV, etc. Layer two has been developed in such a way that, without layer three, it simply translates a code fragments to z architecture code, executes that code, then discards the translated fragment. It detects the interface stub for layer three and, if that is present, it takes advantage of the capabilities including prior translation reuse. Layer 3 is multithreaded and is the cause/source of the problem. Whether layer 3 is run with layers 1 and 2 or in standalone test mode the results are the same. First the environment: VM 4.3 Number of processors: 2 Virtual CPUs (from 2 to 6 .. See note below) Now the application from 10,000 feet: Layer three consists of a parent thread that creates 4 additional threads. Each thread is created in a dispatch class that is unique. Routines are not shared between threads. Upon entry into each routine, the preamble is destroyed and restored on exit to trap any potential inadvertent share. Critical fields shared between threads are protected by a compare and swap spin lock. Part of the testing consists of pushing 1WAY IUCV messages from each connected client every 20 milliseconds. The VM directory for each of 4 machines (one server and three clients) defines the machine as an XC mode machine with: CPU 00 BASE CPU 01 As each thread is created it requests either BASE or ANY CPU affinity. BASE affinity is reserved for the parent and IUCV message handler threads . ANY is used for all other threads . Each affinity request receives a normal return code. All this works beautifully for days and millions of messages UNTIL the number of virtual CPUs defined exceed the number of real CPUs assigned to the VM image. When this takes place, everything comes unstuck. By everything I mean everything in CMS. Stack overflow (03FF abend), free storage management failure, all of it. The multitasking application dev guide states that to the extent possible, dispatch classes are assigned to vCPUs and further states that the max number of vCPUs that may be utilized is equal to the number of dispatch classes. Whether the vCPUs are defined in the user directory entry OR they are created dynamically using the CPU Create CMS function, the results are the same. My questions) 1. Has anyone had a similar experience? 2. Is this a known issue with 4.3? Or in more current releases? 3. Although this seems to be telling me no to go there, I¹ve tried but cannot find anything that says ³You¹ll shoot your eye out , kid.² if you define more virtual CPUs than real processors. Anyone know of such a restriction? 4. Is it possible that CMS kernel services don¹t tolerate a situation where the number of virtual CPUs exceeds ³real² processors? Thanks in advance for any insight you might have on this behavior. --. .- .-. -.-- Gary Dennis Mantissa Corporation
Full volume minidisk FTP
Using VMFTP, how is the best way to transmit full volume minidisks (mixed CMS, VSE, and z/OS) to a backup site? --. .- .-. -.-- Gary Dennis
Re: IUCV - What's wrong with this picture?
Thanks for the response on the IUCV questions. I have included below item 6 from the thread origin and a snippet from John Baker's response. Maybe I should have placed more emphasis on item 6. The server machine is going to be updating the buffer areas in all the connected client machines. Therefore, he server machine needs to know immediately when one of the guests is quiescent or logged off. IUCV will inform the server when a connection is severed. The guest machines can set in indicator in an area monitored by the server to indicate that they have begun a normal closedown *but* the fall-off_the page case is when a machine is logged off and the server attempts to access the buffers in a machine that no longer exists. John made a good argument for temporary IUCV connections. In that case the best way to make a determination on the active state a diagnose that issues a query command for the user in question? On a clear day in the not too distant past, Gary Dennis, [EMAIL PROTECTED] wrote 6. After the initial call, the server virtual machine will maintain the buffer table entries in each client virtual machine without additional IUCV interaction. On 8/25/08 9:32 PM, John P. Baker [EMAIL PROTECTED] wrote: Maintaining thousands of IUCV connections may not have a significant impact on real storage considering the vast amounts of memory now available on zSeries processors. However, searching through all of those linked lists WILL have a performance impact, and is totally unnecessary. --. .- .-. -.-- Gary Dennis
Re: IUCV - What's wrong with this picture?
Alan, Thanks. Especially for 5 through 9. On 8/27/08 12:37 PM, Alan Altmark [EMAIL PROTECTED] wrote: I don't understand why you want to use temporary connections. I don't. The original idea was to have connections for each guest active all the time **so long as the associated overhead would not be excessive**. You mention an area that is monitored by the server. That sounds like shared memory. I'm not sure the area is shared in the sense you mean shared. The guest machines will each have code fragment table(s) in primary address space. The service machine will maintain the tables for each guest after it has been has granted write access. IUCV can handle hundreds of connections with no problem. We think the high-end number of virtual machines could be in the thousands but if the overhead is low and the message incidence is low, it's a go. --. .- .-. -.-- Gary Dennis Mantissa Corporation
IUCV - What's wrong with this picture?
Assumptions: 0. A VM server machine 1. A cluster of client virtual machines (possibly thousands) 2. n buffers are allocated for each client virtual machine 3. Each buffer contains table elements that require (a) Element ageing (b) Element deletion when invalidated by: 1. lack of use 2. client machine request (c) Compression as buffer fragmentation occurs 4. Each client virtual machine in the cluster is connected via IUCV to the server virtual machine. 5. IUCV traffic between the server machine and client machine is extremely low volume. Initial call, termination call, intermittent statistics call. 6. After the initial call, the server virtual machine will maintain the buffer table entries in each client virtual machine without additional IUCV interaction. Now the questions: 1. Does IUCV infrastructure overhead specifically associated with number of connections become prohibitive at some well known point? 2. Has anyone had experience with an application having a high IUCV connection count like this? If so, what was that experience? Again, the traffic incidence per connection is very low but the number of connections is potentially very high. Thanks --. .- .-. -.-- Gary Dennis
OSA Adapter TCP/IP stack association limit?
Did I read somewhere (the where being a place I cannot at this point locate) that the number of IP stacks which could be associated with a single OSA adapter was 640? Running several thousand desktop systems on System z is meaningful only if those operating systems can access a (the) network. Is that stack association limit correct? If this is the case, is there a circumvention? Can multiple OSA adapters be associated with a single VLAN? I'm trying to figure out if this or some other software/hardware constraint is a limiting factor for the number of virtual machines which can share a network on or through System z. --. .- .-. -.-- Gary Dennis Mantissa Corporation
Re: OSA Adapter TCP/IP stack association limit?
Will using VSWITCH get us around the 640 limit per OSA adapter? --. .- .-. -.-- Gary Dennis On 8/6/08 1:35 PM, Mark Post [EMAIL PROTECTED] wrote: On 8/6/2008 at 2:24 PM, in message [EMAIL PROTECTED], Gary M. Dennis [EMAIL PROTECTED] wrote: Did I read somewhere (the where being a place I cannot at this point locate) that the number of IP stacks which could be associated with a single OSA adapter was 640? That sounds right. -snip- Is that stack association limit correct? If this is the case, is there a circumvention? Can multiple OSA adapters be associated with a single VLAN? Use a VSWITCH, since you've already decided this is going to be run on z/VM. Mark Post
Re: z/VM 5.4 Workloads
Can someone point me to the source of the announcement excerpt below? --. .- .-. -.-- Gary Dennis Mantissa Corporation On 8/5/08 11:51 AM, Lionel B. Dyck [EMAIL PROTECTED] wrote: This is an interesting bullet in the announcement: Moving selected Linux, Windows®, and UNIX® workloads to a single System z server: Moving workloads while maintaining distinct server images and current LAN topology can help reduce systems-management complexity. By managing large server farms deployed on virtual servers instead of using multiple hardware servers, the number of real hardware servers and associated physical LANs may be reduced, which may lead to cost savings. Deploying Linux workloads on z/VM V5.4 may be particularly attractive if they interact with applications or data located on the same System z server. Is IBM pre-announcing the ability to run Windows programs on System z? :-) Lionel B. Dyck, Consultant/Specialist Enterprise Platform Services, Mainframe Engineering KP-IT Enterprise Engineering 925-926-5332 (8-473-5332) | E-Mail: [EMAIL PROTECTED] mailto:[EMAIL PROTECTED] AIM: lbdyck | Yahoo IM: lbdyck Kaiser Service Credo: Our cause is health. Our passion is service. We're here to make lives better. I never guess. It is a capital mistake to theorize before one has data. Insensibly one begins to twist facts to suit theories, instead of theories to suit facts. - Sir Arthur Conan Doyle NOTICE TO RECIPIENT: If you are not the intended recipient of this e-mail, you are prohibited from sharing, copying, or otherwise using or disclosing its contents. If you have received this e-mail in error, please notify the sender immediately by reply e-mail and permanently delete this e-mail and any attachments without reading, forwarding or saving them. Thank you.
Re: Nice idea in blog: Should we toss x86 architecture - NOT.
32 and 16 bit. The boot loaders used and FreeDos required incorporation of 16 bit support. Monumental pain. We don't see 64 bit support being problematic though. Gary On 8/1/08 9:31 AM, Adam Thornton [EMAIL PROTECTED] wrote: Does z/VOS do x86_64 or just 32-bit x86? Actually, can you list which x86 extensions are included in it?
Re: Nice idea in blog: Should we toss x86 architecture - NOT.
Apologies for not responding to this thread in a more timely fashion. I had a flood of emails after the initial post. Speed OR Portability Adam is closer than he knows about the approach we have taken on z/VOS. First, he is right when he guessed almost-certainly assembly. We have tried both QEMU and BOCH and you can forget your name before the target OS completes IPL in our current environment. This is not meant to be a criticism. Both of these systems were written for portability rather than platform specific speed. He is also on target about the instruction mapping. Obviously the closer you get to a one to one instruction relationship the better the performance. Pipeline/Instruction Overlap Making the instruction stream pipeline-friendly has been another consideration with z/VOS. David Bond gave a great presentation (Share Session 8192 in August 2006) on this so I leave it to you and Google if you are interested. Single Pass Translation For those of you that think emulation of x86 would be a bad idea, we agree. z/VOS translates guest OS code during initial execution. Code fragment storage, lookup, disposal and reuse for primary and sibling guests are addressed in a patent application. Suffice it to say that we don't interpret or emulate massive amounts of x86 code for use 2-n. z/VOS Development Environment Mantissa has a FLEX development environment that redefines and enhances the definition of slow. We have two copies of VM operational at all times. One of the VM system supports z/OS and VSE development. The other system is dedicated to z/VOS development. Because Adam mentioned FreeDos, I will give this point of reference. Under z/VOS FreeDOS IPLs in 1 second with all debug logging interfaces enabled. After the IPL, it is split second responsive. --. .- .-. -.-- Gary Dennis Mantissa Corporation On 7/25/08 11:28 AM, Adam Thornton [EMAIL PROTECTED] wrote: On Jul 25, 2008, at 8:48 AM, McKown, John wrote: Somewhat like BOCH? I remember somebody saying that they ran Windows on BOCH on an old P/390. A little more data: the straight-up portable-emulation x86 code-path is still not a good idea. I got the current released bochs (20080720) built (with all the cool stuff like x86_64, SSE, plenty of neat features) on Linux s390x. The build was quite clean, actually. It is a lot less painful than on a P/390 or H70, but running Bochs on a 2094 (z9 of some sort), where z/VM sees 2 CPUs but the Linux guest has one, is still only giving me 3.4 to 4.0 million (x86) instructions per second, which is...well, a LOT less than you'd get on a modern Xeon. That's not to say that I'm necessarily CPU-bound. If I had time to play with it, the VGA refresh rate is where I'd start, because that probably isn't helping. FreeDOS installation was pokey but not really terrible. Performance is, well pretty bad; it feels like working over a satellite link in terms of latency. I think you'd have a really hard time making the case to management that THIS was a good use for your zSeries. So here's hoping that the Mantissa product is focussed around an efficient (and almost-certainly assembly) x86 emulation. Given the richness of the s390x instruction set, and that a bunch of the instructions fundamentally do the same thing in the x86 and the s390x world (that is, move something from a memory location to a named register is the same concept on either architecture), I would hope that most of the user-mode instructions can be mapped close to 1-to-1, and the mere fact of having to create an instruction translator is going to mean that the actual performance will be several-host- instructions-to-one-guest-instruction. Complicated instructions are still going to be slowed significantly, of course. Adam --. .- .-. -.-- Gary Dennis Mantissa Corporation 1121 Edenton Street Birmingham, Alabama 35242-9257 p: 205.968-3942 m: 205.218-3937 f: 205.968.3932 [EMAIL PROTECTED] http://www.mantissa.com http://www.idovos.com
Re: Nice idea in blog: Should we toss x86 architecture - NOT.
z/VOS is written to support the x86 instruction set and the underlying hardware rather than a specific operating system. For example, FreeDos was used as the initial debug target operating system due to source code availability. --. .- .-. -.-- Gary Dennis Mantissa Corporation On 7/23/08 9:06 AM, Mary Anne Matyaz [EMAIL PROTECTED] wrote: Gary, if it runs native windows, will it also then run x86 linux? That seems to be one of the barriers for us, that z/linux may not support certain x86 linux applications. Thanks, Mary Anne Gary M. Dennis wrote: Z/VOS is a CMS application. The glass-side user will only see Windows via RDC and know nothing of or about CMS or VM. Gary On 7/22/08 8:30 PM, dave [EMAIL PROTECTED] wrote: Good luck, Gary. I do hope your organization can pull this off. VM-ers need more employment possibilities:-) I gather from some of your previous posts to this list that your Windows support software, z/VOS, is in fact a sophisticated CMS-based application, that is a user would log onto a CMS user id to start his Windows systemis my understanding correct? Thanks and have a good one. DJ - Original Message - From: Gary M. Dennis [EMAIL PROTECTED] To: IBMVM@LISTSERV.UARK.EDU Subject: Re: Nice idea in blog: Should we toss x86 architecture Date: Tue, 22 Jul 2008 13:02:33 -0500 This was our post to the zd net blog. Maybe we already have. In Q1 2009 Mantissa will deliver a system that permits unaltered Windows operating systems to run under z/VM. Using a desktop appliance running RDC, users will be able to connect to their virtual Windows images running in the VM environment. Goodbye desktop hardware, remote maintenance, high power consumption, machine order lead time. z/VOS began with the observation that most Windows workstations do practically nothing 95% of the time and we were so intrigued with the idea of being able to actually run an intel-based operating system under IBM VM that we never looked back. VM provided a natural platform for development of this product. The product has been a bear for the development group but the thought of being able to run 3000 copies of Windows on one System z so fascinated the team that we needed very little additional incentive. Let's hope IBM can ramp up System z production. Why wait until 2016? --. .- .-. -.-- Gary Dennis Mantissa Corporation On 7/22/08 11:14 AM, Bob Heerdink [EMAIL PROTECTED] wrote: http://blogs.zdnet.com/perlow/?p=9183 Should we toss x86 architecture and wipe the slate with something greene r and more scalable? Windows Server 2016 128-bit edition running virtualized on z/VM in a gre en datacenter, accessed via my house from a thin client over high-speed fibe r optic connection. I can see it now. Hope this happens sooner than predicted, Bob
Re: Nice idea in blog: Should we toss x86 architecture
This was our post to the zd net blog. Maybe we already have. In Q1 2009 Mantissa will deliver a system that permits unaltered Windows operating systems to run under z/VM. Using a desktop appliance running RDC, users will be able to connect to their virtual Windows images running in the VM environment. Goodbye desktop hardware, remote maintenance, high power consumption, machine order lead time. z/VOS began with the observation that most Windows workstations do practically nothing 95% of the time and we were so intrigued with the idea of being able to actually run an intel-based operating system under IBM VM that we never looked back. VM provided a natural platform for development of this product. The product has been a bear for the development group but the thought of being able to run 3000 copies of Windows on one System z so fascinated the team that we needed very little additional incentive. Let's hope IBM can ramp up System z production. Why wait until 2016? --. .- .-. -.-- Gary Dennis Mantissa Corporation On 7/22/08 11:14 AM, Bob Heerdink [EMAIL PROTECTED] wrote: http://blogs.zdnet.com/perlow/?p=9183 Should we toss x86 architecture and wipe the slate with something greene r and more scalable? Windows Server 2016 128-bit edition running virtualized on z/VM in a gre en datacenter, accessed via my house from a thin client over high-speed fibe r optic connection. I can see it now. Hope this happens sooner than predicted, Bob
Re: Nice idea in blog: Should we toss x86 architecture
We looked very hard at the licensing aspect of this. We don't see anything in the Microsoft EULA that would permit or cause them to treat this environment different any different than existing VM environments. This environment should work in their favor since the images (and therefore the licenses) can be deployed more efficiently than in an blade warehouse environment. --. .- .-. -.-- Gary Dennis Mantissa Corporation On 7/22/08 1:07 PM, Schuh, Richard [EMAIL PROTECTED] wrote: Interesting idea. What are the licensing issues? (And is VM Vista Ready? ;-) ) Regards, Richard Schuh -Original Message- From: The IBM z/VM Operating System [mailto:[EMAIL PROTECTED] On Behalf Of Gary M. Dennis Sent: Tuesday, July 22, 2008 11:03 AM To: IBMVM@LISTSERV.UARK.EDU Subject: Re: Nice idea in blog: Should we toss x86 architecture This was our post to the zd net blog. Maybe we already have. In Q1 2009 Mantissa will deliver a system that permits unaltered Windows operating systems to run under z/VM. Using a desktop appliance running RDC, users will be able to connect to their virtual Windows images running in the VM environment. Goodbye desktop hardware, remote maintenance, high power consumption, machine order lead time. z/VOS began with the observation that most Windows workstations do practically nothing 95% of the time and we were so intrigued with the idea of being able to actually run an intel-based operating system under IBM VM that we never looked back. VM provided a natural platform for development of this product. The product has been a bear for the development group but the thought of being able to run 3000 copies of Windows on one System z so fascinated the team that we needed very little additional incentive. Let's hope IBM can ramp up System z production. Why wait until 2016? --. .- .-. -.-- Gary Dennis Mantissa Corporation On 7/22/08 11:14 AM, Bob Heerdink [EMAIL PROTECTED] wrote: http://blogs.zdnet.com/perlow/?p=9183 Should we toss x86 architecture and wipe the slate with something greene r and more scalable? Windows Server 2016 128-bit edition running virtualized on z/VM in a gre en datacenter, accessed via my house from a thin client over high-speed fibe r optic connection. I can see it now. Hope this happens sooner than predicted, Bob
Re: Nice idea in blog: Should we toss x86 architecture
No web site exists for z/VOS. The target for the web site is product launch date -45 (November 15). --. .- .-. -.-- Gary Dennis Mantissa Corporation On 7/22/08 1:07 PM, McKown, John [EMAIL PROTECTED] wrote: -Original Message- From: The IBM z/VM Operating System [mailto:[EMAIL PROTECTED] On Behalf Of Gary M. Dennis Sent: Tuesday, July 22, 2008 1:03 PM To: IBMVM@LISTSERV.UARK.EDU Subject: Re: Nice idea in blog: Should we toss x86 architecture This was our post to the zd net blog. Maybe we already have. In Q1 2009 Mantissa will deliver a system that permits unaltered Windows operating systems to run under z/VM. Using a desktop appliance running RDC, users will be able to connect to their virtual Windows images running in the VM environment. Goodbye desktop hardware, remote maintenance, high power consumption, machine order lead time. Is there a web page on this that I could relay to my manager? -- John McKown Senior Systems Programmer HealthMarkets Keeping the Promise of Affordable Coverage Administrative Services Group Information Technology The information contained in this e-mail message may be privileged and/or confidential. It is for intended addressee(s) only. If you are not the intended recipient, you are hereby notified that any disclosure, reproduction, distribution or other use of this communication is strictly prohibited and could, in certain circumstances, be a criminal offense. If you have received this e-mail in error, please notify the sender by reply and delete this message without copying or disclosing it. --. .- .-. -.-- Gary Dennis Mantissa Corporation
Re: Nice idea in blog: Should we toss x86 architecture
When I read this post I sat at my machine for about 2 minutes trying to figure out a good way to respond to your question. In that 2 minutes nothing on my screen changed except the seconds field on the time box. Everything else in the video buffer stayed absolutely the same during that wait. It is highly unlikely in a business environment that everyone would run a compute intensive application such as a screen saver to run on 3000 virtual machines simultaneously. A screen screen saver would run on the appliance that supported the RDC environment (and that actually had a screen to save). Again, most desktop machines don't actually do enough from a task activity and graphics standpoint to create overwhelming overhead. --. .- .-. -.-- Gary Dennis Mantissa Corporation On 7/22/08 1:44 PM, Romanowski, John (OFT) [EMAIL PROTECTED] wrote: How can the z handle 3000 copies of Windows all running a graphic user interface (cpu-intensive) ? This e-mail, including any attachments, may be confidential, privileged or otherwise legally protected. It is intended only for the addressee. If you received this e-mail in error or from someone who was not authorized to send it to you, do not disseminate, copy or otherwise use this e-mail or its attachments. Please notify the sender immediately by reply e-mail and delete the e-mail from your system. -Original Message- From: The IBM z/VM Operating System [mailto:[EMAIL PROTECTED] On Behalf Of Gary M. Dennis Sent: Tuesday, July 22, 2008 2:03 PM To: IBMVM@LISTSERV.UARK.EDU Subject: Re: Nice idea in blog: Should we toss x86 architecture This was our post to the zd net blog. Maybe we already have. In Q1 2009 Mantissa will deliver a system that permits unaltered Windows operating systems to run under z/VM. Using a desktop appliance running RDC, users will be able to connect to their virtual Windows images running in the VM environment. Goodbye desktop hardware, remote maintenance, high power consumption, machine order lead time. z/VOS began with the observation that most Windows workstations do practically nothing 95% of the time and we were so intrigued with the idea of being able to actually run an intel-based operating system under IBM VM that we never looked back. VM provided a natural platform for development of this product. The product has been a bear for the development group but the thought of being able to run 3000 copies of Windows on one System z so fascinated the team that we needed very little additional incentive. Let's hope IBM can ramp up System z production. Why wait until 2016? --. .- .-. -.-- Gary Dennis Mantissa Corporation On 7/22/08 11:14 AM, Bob Heerdink [EMAIL PROTECTED] wrote: http://blogs.zdnet.com/perlow/?p=9183 Should we toss x86 architecture and wipe the slate with something greene r and more scalable? Windows Server 2016 128-bit edition running virtualized on z/VM in a gre en datacenter, accessed via my house from a thin client over high-speed fibe r optic connection. I can see it now. Hope this happens sooner than predicted, Bob
Re: Nice idea in blog: Should we toss x86 architecture
Z/VOS is a CMS application. The glass-side user will only see Windows via RDC and know nothing of or about CMS or VM. Gary On 7/22/08 8:30 PM, dave [EMAIL PROTECTED] wrote: Good luck, Gary. I do hope your organization can pull this off. VM-ers need more employment possibilities:-) I gather from some of your previous posts to this list that your Windows support software, z/VOS, is in fact a sophisticated CMS-based application, that is a user would log onto a CMS user id to start his Windows systemis my understanding correct? Thanks and have a good one. DJ - Original Message - From: Gary M. Dennis [EMAIL PROTECTED] To: IBMVM@LISTSERV.UARK.EDU Subject: Re: Nice idea in blog: Should we toss x86 architecture Date: Tue, 22 Jul 2008 13:02:33 -0500 This was our post to the zd net blog. Maybe we already have. In Q1 2009 Mantissa will deliver a system that permits unaltered Windows operating systems to run under z/VM. Using a desktop appliance running RDC, users will be able to connect to their virtual Windows images running in the VM environment. Goodbye desktop hardware, remote maintenance, high power consumption, machine order lead time. z/VOS began with the observation that most Windows workstations do practically nothing 95% of the time and we were so intrigued with the idea of being able to actually run an intel-based operating system under IBM VM that we never looked back. VM provided a natural platform for development of this product. The product has been a bear for the development group but the thought of being able to run 3000 copies of Windows on one System z so fascinated the team that we needed very little additional incentive. Let's hope IBM can ramp up System z production. Why wait until 2016? --. .- .-. -.-- Gary Dennis Mantissa Corporation On 7/22/08 11:14 AM, Bob Heerdink [EMAIL PROTECTED] wrote: http://blogs.zdnet.com/perlow/?p=9183 Should we toss x86 architecture and wipe the slate with something greene r and more scalable? Windows Server 2016 128-bit edition running virtualized on z/VM in a gre en datacenter, accessed via my house from a thin client over high-speed fibe r optic connection. I can see it now. Hope this happens sooner than predicted, Bob
CMS Multi-tasking - How is it different from z/OS task management
We are attempting to leverage z/VM CMS multi-tasking capabilities for Windows® thread management. We have substantial z/OS experience with regard to task management and serialization so what we are experiencing in a CMS environment doesn¹t fit what we had anticipated. Our test was conducted on a virtual machine for which two CPU¹s are defined. The test program contains CSECTS MAIN and THREAD1. The program is invoked under CMS. CSECT MAIN process 1. Initialization 2. Call VM thread create (referencing THREAD1 CSECT) from within MAIN CSECT in order to create THREAD1 process in a different class (new-Class specified on thread create call). 3. Issue WTO repeatedly THREAD1 CSECT process 1. Initialization 2. Issue WTO repeatedly What we expected: Interspersed WTO's from both MAIN and THREAD1 threads What we get: THREAD1 WTOs only. We thought MAIN, being in a separate class (and therefore eligible to be assigned to different CPUs) would dispatch (and both issue WTOs) but this did not happen. In z/OS the fact that the WTO was issued would provide sufficient dispatch latency for another task to get a time slice. Additional observations: 1. If line write is substituted for WTO in MAIN and THREAD1 there is no observed difference. 2. The program works as expected IF yield is called within the WTO loops in MAIN and THREAD1. Why should yield have to be called? Any thread wizards out there? --. .- .-. -.-- Gary Dennis Mantissa Corporation
VM - Network best practices
We need to put together something approaching a production network environment for Windows® under z/VM testing. We don't believe a 500 seat environment would generate any more network traffic or for that matter be any more complex than the network definitions for a z/VM Linux server colony. Has anyone put together a fairly complex multi-guest VM network using VSWITCH? If so, can you point me to any VM definitions that may have been shared on this list? Which IBM PUB is a definitive cookbook on this? Thanks --. .- .-. -.-- Gary Dennis Mantissa
Re: Question about z/VM...
On 4/10/08 10:02 AM, David Boyes [EMAIL PROTECTED] wrote: You cannot run Intel binaries efficiently on System z hardware. You can run Windows applications based on the portable subset of .NET (using Mono) or applications which you have source code and can recompile for System z hardware. There are also suites that allow some ASP applications to run. Pure Java should run if you have the right combination of JVM and environment. The major Java container applications (WAS, BEA, jboss/tomcat) run well (with a bit of tuning). Our experience with z/VOS indicates that it is possible to run Intel binaries efficiently on System z. However, this efficiency is not possible with emulation. Prime pass code translation with managed code segment invalidation is the only way we have found to achieve performance viability using native binaries. Many performance efficiencies not apparent initially became obvious once we started to reconcile the IBM and Intel principles of operation. You technically can run Windows in a z/VM virtual machine using a Intel emulator like bochs, but the overhead CPU cost is horrendous (75-100 to 1). You wouldn¹t want to do it for production work unless you have lots of money to burn, but it might be OK for testing stuff. We have run Windows® (98/NT) under both BOCHS and QEMU on System z under Cent OS and can attest to the overhead and cost. This nightmare experience is the reason we selected z/VM CMS as the environment for z/VOS. The approach taken by QEMU and BOCHS ensures that neither system will achieve viable performance/resource consumption metrics on System z. In defense of these products, they were never intended to do this. A z/VM runtime environment won't improve guest reliability but if that guest can be run at a fraction of the cost, the value proposition would be compelling. If 25 virtualized images on one box is a good thing, 500 guests on a single footprint would be attractive. There could be a major change in where and how computing resources are consumed. If you think that type of shift could take place (it has already has with intel virtualization systems), the possibility that the source of those cycles may be different is not a stretch. If you need dense numbers Windows servers, look at the newest quad and octo-core blade servers with lots and lots of RAM running VMWare. They¹re about the best available option for the typical Windows application server sprawl. You should look at whether some of your Intel Linux apps could be moved, though, or pieces of infrastructure like Oracle or DB/2 servers could be moved. There are substantial savings to be had in terms of licensing for infrastructure pieces. --. .- .-. -.-- Gary Dennis
Re: Question about z/VM...
z/VOS has not been released and no technical specifications are available. It should not surprise anyone that this could and would be done. It's much closer to detective work than rocket science. Once you acknowledge that the x86 instruction set is a jungle when compared to the well ordered garden to which most of us are accustomed, building the instruction support is tedious but straight forward. The challenge is not getting this to work; it's getting it to work well. --. .- .-. -.-- Gary Dennis On 4/10/08 4:25 PM, Patrick Spinler [EMAIL PROTECTED] wrote: -BEGIN PGP SIGNED MESSAGE- Hash: SHA1 Gary M. Dennis wrote: | | Our experience with z/VOS indicates that it is possible to run Intel | binaries efficiently on System z. However, this efficiency is not possible | with emulation. Prime pass code translation with managed code segment | invalidation is the only way we have found to achieve performance viability | using native binaries. Many performance efficiencies not apparent initially | became obvious once we started to reconcile the IBM and Intel principles of | operation. | (snip) | | We have run Windows® (98/NT) under both BOCHS and QEMU on System z under | Cent OS and can attest to the overhead and cost. This nightmare experience | is the reason we selected z/VM CMS as the environment for z/VOS. | Where might one find more information about the z/VOS product you mention? A quick google search or two doesn't show any obvious hits on the first pages .. - -- Pat -BEGIN PGP SIGNATURE- Version: GnuPG v1.4.7 (Darwin) Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org iD8DBQFH/oXhNObCqA8uBswRAtmQAJ9CIxJJV4G34C6tAV0hmhaIsCsdYgCfazA+ WG8sGvRGZBvj37HY/kOMkk0= =gZrV -END PGP SIGNATURE-
Re: z/VM - Lightweight specific purpose file system
On 3/26/08 5:05 PM, Dave Wade [EMAIL PROTECTED] wrote: The existing licenses already allow running in a virtual environment and don't specify what chips etc that could be. They could change future licenses, perhaps, but MS licenses don't work like Mainframe Licenses and it would be hard to exclude mainframe based emulation without excluding VM Ware. Since z/VOS is neither emulation or paravirtualization it is conceivable that an attorney might take exception to how the MS EULA applied to running Windows/XP in the z/VM environment. Because Apple has been reluctant to take a clear stance on virtualization of their products we sent their legal department a letter asking for clarification of their position on the issue of running OS X under z/VM and received no response. They may still be working their way through the letter using Wikipedia to decode some of the acronyms. Ultimately we don't think either company will challenge the product on the basis of hardware platform. Here's why. 1. Its deep Green. 2. It's efficient and TCO positive 3. License sales will, in all likelihood, not go down. I guess they could buy VM Ware first... If Microsoft waits until after the release of this product they maybe able to buy VM Ware for substantially less. Gary Dennis Mantissa Corporation Dave G4UGM Illegitimi Non Carborundum - Original Message - From: McKown, John [EMAIL PROTECTED] To: IBMVM@LISTSERV.UARK.EDU Sent: Wednesday, March 26, 2008 7:09 PM Subject: Re: z/VM - Lightweight specific purpose file system -Original Message- From: The IBM z/VM Operating System [mailto:[EMAIL PROTECTED] On Behalf Of Dave Wade Sent: Wednesday, March 26, 2008 2:01 PM To: IBMVM@LISTSERV.UARK.EDU Subject: Re: z/VM - Lightweight specific purpose file system Why would the Microsoft Licensing be tricky? Expensive perhaps as you need one license per virtual machine, but not tricky... Well, tricky in that MS might refuse to grant the license. They are under no obligation to do so. And they are really, really worried about Windows under any virtualization other than their own. Running on unsupported hardware would likely make them even more reluctant. Of course, I cannot think of any software that runs on Windows that I would want to run on a z. I'd rather replace any such with equivalent software, if there is some, or just run on Intel for that function. -- John McKown Senior Systems Programmer HealthMarkets Keeping the Promise of Affordable Coverage Administrative Services Group Information Technology The information contained in this e-mail message may be privileged and/or confidential. It is for intended addressee(s) only. If you are not the intended recipient, you are hereby notified that any disclosure, reproduction, distribution or other use of this communication is strictly prohibited and could, in certain circumstances, be a criminal offense. If you have received this e-mail in error, please notify the sender by reply and delete this message without copying or disclosing it. --. .- .-. -.-- Gary Dennis Mantissa Corporation
Re: z/VM - Lightweight specific purpose file system
Early in the development cycle, we had both QEMU and Bochs running on z/System version of Redhat (CentOS 5.4). The Name two movie stars and a dog joke applied to both emulators running in this environment. We concluded early on that we had to get rid of Linux and the emulation layer. Both would prevent us from ever achieving the required level of performance. The result of that detour is that the only thing between Windows® and VM is CMS and translation code. On 3/26/08 9:57 AM, Adam Thornton [EMAIL PROTECTED] wrote: An excellent goal. As a point of comparison, have you ever run Windows using the Bochs emulator on zLinux? If so, on what machine? (I'd like to see someone try it on a z10.) --. .- .-. -.-- Gary Dennis Mantissa Corporation 2 Perimeter Park South Birmingham, Alabama 35243-3274 p: 205.968-3942 m: 205.218-3937 f: 205.968.3932 [EMAIL PROTECTED] http://www.mantissa.com http://www.idovos.com
z/VM - Lightweight specific purpose file system
We need a lightweight file system to support z/VM i86 guest operating systems. A high speed garbage can of sorts. Is anyone aware of a VM open source file system port with some of the characteristics listed below. Such a system might enable us to add the functionality needed to support these guests without starting at zero. 1. Large file allocation capability; potentially multi-terabyte. A file in this instance represents a drive in the PC world (the boot disk for instance; C: drive). The system needs to support a few thousand very large files rather than millions of small files. 2. Allocation and access for variable interval sizes (A multiple of 512 up to 4096) 3. Sparse allocation. For example, an allocation of 250 GB would create the required structure for 250 GB but not actually allocate the storage intervals until required. 4. Support for very large shared storage pools Thanks Gary Dennis Mantissa
Re: z/VM - Lightweight specific purpose file system
The callable services benchmarks we conducted with BFS ran between 8 and 10 times longer than the test set running with the CMS file system. Assuming a cluster of 125 Windows® 2K z/VM guests and using I/O counts generated by Win 2K on native Intel hardware the results of extrapolating the I/O overhead spooked us a bit. In effect, all our instruction pipeline optimization and translated instruction segment reuse optimization would be negated by the I/O overhead. We have a callable file system for z/OS that can handle an array of 128 pools each containing up to 255 volumes each. That system would be a bear to convert owing to the OS-specific interface code but it appears from your comments that converting may have to be seriously considered to achieve the desired results. Thank you. Gary Dennis Mantissa On 3/25/08 9:55 AM, Alan Altmark [EMAIL PROTECTED] wrote: On Tuesday, 03/25/2008 at 04:26 EDT, Gary M. Dennis [EMAIL PROTECTED] wrote: Is anyone aware of a VM open source file system port with some of the characteristics listed below. Such a system might enable us to add the functionality needed to support these guests without starting at zero. It isn't Open Source, but CMS has a POSIX file system (Byte File System, BFS) that is managed by the SFS server, allocating space only as used. I don't know that I would classify it as lightweight, though from the CMS user's point of view, it is, since the I/O takes place in the SFS server, but it takes APPC/VM (IUCV on steriods) calls to make it happen. You can talk to it in assembler using the BPX1 callable services. It could provide you a jump start while you develop your own file system. And just in case you haven't discovered it already, there's no pluggable file system interface in CMS. You will need to write your file system from the bottom up. The only help CMS will provide to you is in the form of HNDIO,HNDSVC, NUCEXT, and NUCXLOAD. Alan Altmark z/VM Development IBM Endicott
Re: z/VM - Lightweight specific purpose file system
Months ago. The development team was so focused on instruction result fidelity, machine state, and segment translation bypass issues that I/O subsystem did not receive the necessary attention. At least the tough part is done. Gary Dennis Mantissa On 3/25/08 4:17 PM, [EMAIL PROTECTED] [EMAIL PROTECTED] wrote: Ummm, I may have missed something, but since when can you run Windows on an IBM mainframe? Peter -Original Message- From: The IBM z/VM Operating System [mailto:[EMAIL PROTECTED] On Behalf Of Gary M. Dennis Sent: March 25, 2008 17:14 To: IBMVM@LISTSERV.UARK.EDU Subject: Re: z/VM - Lightweight specific purpose file system The callable services benchmarks we conducted with BFS ran between 8 and 10 times longer than the test set running with the CMS file system. Assuming a cluster of 125 Windows(r) 2K z/VM guests and using I/O counts generated by Win 2K on native Intel hardware the results of extrapolating the I/O overhead spooked us a bit. In effect, all our instruction pipeline optimization and translated instruction segment reuse optimization would be negated by the I/O overhead. We have a callable file system for z/OS that can handle an array of 128 pools each containing up to 255 volumes each. That system would be a bear to convert owing to the OS-specific interface code but it appears from your comments that converting may have to be seriously considered to achieve the desired results. Thank you. Gary Dennis Mantissa On 3/25/08 9:55 AM, Alan Altmark [EMAIL PROTECTED] wrote: On Tuesday, 03/25/2008 at 04:26 EDT, Gary M. Dennis [EMAIL PROTECTED] wrote: Is anyone aware of a VM open source file system port with some of the characteristics listed below. Such a system might enable us to add the functionality needed to support these guests without starting at zero. It isn't Open Source, but CMS has a POSIX file system (Byte File System, BFS) that is managed by the SFS server, allocating space only as used. I don't know that I would classify it as lightweight, though from the CMS user's point of view, it is, since the I/O takes place in the SFS server, but it takes APPC/VM (IUCV on steriods) calls to make it happen. You can talk to it in assembler using the BPX1 callable services. It could provide you a jump start while you develop your own file system. And just in case you haven't discovered it already, there's no pluggable file system interface in CMS. You will need to write your file system from the bottom up. The only help CMS will provide to you is in the form of HNDIO,HNDSVC, NUCEXT, and NUCXLOAD. Alan Altmark z/VM Development IBM Endicott The information transmitted is intended only for the person or entity to which it is addressed and may contain confidential and/or privileged material. Any review retransmission dissemination or other use of or taking any action in reliance upon this information by persons or entities other than the intended recipient or delegate is strictly prohibited. If you received this in error please contact the sender and delete the material from any computer. The integrity and security of this message cannot be guaranteed on the Internet. The sender accepts no liability for the content of this e-mail or for the consequences of any actions taken on the basis of information provided. The recipient should check this e-mail and any attachments for the presence of viruses. The sender accepts no liability for any damage caused by any virus transmitted by this e-mail. This disclaimer is property of the TTC and must not be altered or circumvented in any manner.
Re: z/VM - Lightweight specific purpose file system
Emulation would be a non-starter for a production environment. I would describe this system as a single pass code segment translation system with conditional block invalidation. We have been using VM for 20 of our 27 years in business. A development environment without it has never been considered an option. Many companies (ours included) consider running a few dozen virtual Windows® images on a rack-mounted machine good business. We see no reason why z/System should not support from 250 images on the low end to several thousand on mid and high end systems. On 3/25/08 5:45 PM, Stephen Frazier [EMAIL PROTECTED] wrote: Are you attempting to write a windows emulator that runs under VM? Looking at your companies web site it looks like you mostly sell products that run under z/OS. If you can do this there will be a lot of interest. Gary M. Dennis wrote: Months ago. The development team was so focused on instruction result fidelity, machine state, and segment translation bypass issues that I/O subsystem did not receive the necessary attention. At least the tough part is done. Gary Dennis Mantissa --. .- .-. -.-- Gary M. Dennis Mantissa Corporation 2 Perimeter Park South Birmingham, Alabama 35243-3274 p: 205.968-3942 m: 205.218-3937 f: 205.968.3932 [EMAIL PROTECTED] http://www.mantissa.com http://www.idovos.com