ty, David, for the interesting point of view, but it certainly does conflict with the comparison numbers IBM showed at the zEnterprise Summit.
IBM, Timothly/Alan, what say ye? On Wed, May 9, 2012 at 12:29 AM, David Crayford <[email protected]> wrote: > On 5/05/2012 2:55 AM, George Henke wrote: > >> tyvm, John, Mark, Edward. >> >> Mark, >> >> Do I need an Enterprise Class z114 box or will a Business Class one >> suffice? >> >> John, >> >> A compelling reason for server consolidation on zBx as IBM pointed out in >> their z Summit is that zMIPS GCPs are totally dedicated to regular >> workload >> processing, no I/O - that is done by the channels, CHPIDs. >> >> Whereas all other servers use the GCPs for both normal processing and I/O. >> So all MIPS are not equal. >> > > What IBM didn't mention in the Z Summit was that offloading I/O to > peripheral hardware hasn't been > exclusive to mainframes for a very long time. Let's take the platform you > want to replace (SPARC) as an example. > > http://developers.sun.com/**solaris/developer/support/** > driver/wps/pci/html/Sun_SPARC.**doc.html<http://developers.sun.com/solaris/developer/support/driver/wps/pci/html/Sun_SPARC.doc.html> > > The terminology may be different but it sure looks like all the I/O, > including interrupts, is offloaded to peripheral processors. > Even the much maligned (on this forum) Intel x86 has DMA and the new E5 > range even has integrated 10GBe and PCIe-3 on the board. > Mainframe didn't get PCIe-2 until last year, which it re-branded FICON > Express8S. > > Years ago when my wife was working for HDS and I had a lively discussion > at a Christmas lunch with some of her colleagues about the the mainframes > superior I/O. They scoffed at me like I had been living in a cave for the > last decade. I was respectfully informed that the high-end UNIX boxes > easily kept pace with the mainframes wrt I/O performance. Most > embarrassing was the fact that mainframes were still huffing and puffing > away on half-duplex ESCON when the competition was racing away on super > fast fibre channel. That was a very humbling experience for me. > > > Since most commerical workloads are I/O, not CPU, bound this amounts to >> substantial CPU savings. >> >> Also, it is the IBM operating system architecture, FLIH, that enables this >> to happen and there is no other server in the world that is configured >> thus. >> > > Maybe I'm missing the point but what does a z/OS FLIH have to do with > porting workloads to zLinux? > > > On Fri, May 4, 2012 at 2:39 PM, McKown, John >> <[email protected]**>wrote: >> >> -----Original Message----- >>>> From: IBM Mainframe Discussion List >>>> [mailto:[email protected]] On Behalf Of George Henke >>>> Sent: Friday, May 04, 2012 1:02 PM >>>> To: [email protected] >>>> Subject: It's feeding time in Jurassic Park . . . >>>> >>>> and the dinosaurs are very hungry. >>>> >>>> I need to migrate 50 Solaris servers to zLinux under z/VM on >>>> a z114 and >>>> about the same number of Windows servers to zBx. >>>> >>> Congrats! >>> >>> Does anyone have experience with this, some ideas? >>>> >>> Experience, no. Ideas, sure! Why not?<grin> >>> >>> Need a migration path. >>>> >>>> Some questions: >>>> >>>> Can z114 be upgraded to include zBx or must I upgrade to a >>>> z196 for that? >>>> >>> I am not sure. >>> >>> Can zBx be SYSPLEXed between 2 CECs? >>>> >>> No. a zBx can be attached to at most 1 CEC. >>> >>> If not, how do I eliminate the SPF, how do I do failover for the blade >>>> servers? >>>> >>> SPF? Single point of failure? You need to have multiple z114s with and >>> associated zBX. >>> I think you'd do failover just like you would with any other "racked >>> mounted" servers. >>> If the rack fails, all the servers fail. So you need the same techniques >>> with a zBX "rack". >>> I'm not a PC person, but I think this involves TCP connections between a >>> server and it's failover >>> twin. >>> >>> How do I convert Solaris to zLInux under z/VM? Solaris is >>>> not supported by >>>> zBx. >>>> Must I recompile the applications running under Solaris to >>>> run them under >>>> zLinux? >>>> >>> Yes, you must recompile. z/Linux runs on the z/Architecture instruction >>> set. Solaris runs on >>> Intel or SPARC(?). You can't run Intel or SPARC instructions on a z. >>> However there are two exceptions >>> that I can think of which __may__ not need recompilation. The first is >>> Java code. The .jar and .class >>> files should run on any compatible JVM, regardless of the underlying >>> hardware architecture. Another >>> possibility is .NET applications. They __might__ run on a z/Linux using >>> Mono. Again, because they >>> don't run "native", but on an instruction emulator. Shell scripts may >>> need >>> some changing. Perl, Python, Ruby >>> and other intepreted language may need some changing, depending on the >>> levels of the intepreter >>> on z/Linux versus the one on Solaris. >>> >>> >>>> -- >>>> George Henke >>>> (C) 845 401 5614 >>>> >>> I wish we were doing something like this. But we're z/OS only and IT >>> management would like to go MS-Windows only. >>> >>> -- >>> John McKown >>> Systems Engineer IV >>> IT >>> >>> Administrative Services Group >>> >>> HealthMarkets(r) >>> >>> 9151 Boulevard 26 * N. Richland Hills * TX 76010 >>> (817) 255-3225 phone * >>> [email protected] * www.HealthMarkets.com >>> >>> Confidentiality Notice: This e-mail message may contain confidential or >>> proprietary information. If you are not the intended recipient, please >>> contact the sender by reply e-mail and destroy all copies of the original >>> message. HealthMarkets(r) is the brand name for products underwritten and >>> issued by the insurance subsidiaries of HealthMarkets, Inc. -The >>> Chesapeake >>> Life Insurance Company(r), Mid-West National Life Insurance Company of >>> TennesseeSM and The MEGA Life and Health Insurance Company.SM >>> >>> >>> >>> ------------------------------**------------------------------** >>> ---------- >>> For IBM-MAIN subscribe / signoff / archive access instructions, >>> send email to [email protected] with the message: INFO IBM-MAIN >>> >>> >> >> > ------------------------------**------------------------------**---------- > For IBM-MAIN subscribe / signoff / archive access instructions, > send email to [email protected] with the message: INFO IBM-MAIN > -- George Henke (C) 845 401 5614 ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO IBM-MAIN
