Re: z990 multiprocessor overhead issue.
On Fri, 2006-03-24 at 05:36 -0600, Patrick Loftus wrote: We seem to be experiencing noticeably less MIPS than expected on our z/OS LPARs, and it almost looks like the speciality engines (zAAP, IFL etc) are contributing to the multiproccesor overheads and reducing the CPU resource. Hadn't thought about that. Would be surprised if IFLs got involved, but I can imagine zAAPs (and zIIPs) might. They are exposed to the MVS scheduler, and so would conceivably suffer the same interprocess locking/spin issues. That's MP factor defined (o.k. just a *little* simplistic). Shane ... -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: z990 multiprocessor overhead issue.
Hadn't thought about that. Would be surprised if IFLs got involved, but I can imagine zAAPs (and zIIPs) might. They are exposed to the MVS scheduler, and so would conceivably suffer the same interprocess locking/spin issues. I think it's more than lock/spin. What about just making the decision to move it to a zIIP, zAAP, or zOWIE? Especially for short transactions? - -teD I’m an enthusiastic proselytiser of the universal panacea I believe in! -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: z990 multiprocessor overhead issue.
With the CPs being pooled, I doubt the specialties draw down the MIP MP effect. They might chew some quick cycles, but they should not impact MIP capacity like additional general purpose CPs would. Bob Richards VP, Enterprise Technologist Enterprise Technology Infrastructure SunTrust Banks, Inc. (404) 575-2798 Seeing beyond money (sm) -Original Message- From: IBM Mainframe Discussion List [mailto:[EMAIL PROTECTED] On Behalf Of Shane Sent: Friday, March 24, 2006 6:21 PM To: IBM-MAIN@BAMA.UA.EDU Subject:Re: z990 multiprocessor overhead issue. On Fri, 2006-03-24 at 00:00 +, Ted MacNEIL wrote: Hadn't thought about that. Would be surprised if IFLs got involved, but I can imagine zAAPs (and zIIPs) might. They are exposed to the MVS scheduler, and so would conceivably suffer the same interprocess locking/spin issues. I think it's more than lock/spin. What about just making the decision to move it to a zIIP, zAAP, or zOWIE? Especially for short transactions? Guess I should have said inter-processor. Looks like we've interpreted the initial post differently. I was thinking of a drop in SU/sec delivery, rather than a loss usable MIPs within an LPAR (capture ratio change ???). LEGAL DISCLAIMER The information transmitted is intended solely for the individual 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 action in reliance upon this information by persons or entities other than the intended recipient is prohibited. If you have received this email in error please contact the sender and delete the material from any computer. Seeing Beyond Money is a service mark of SunTrust Banks, Inc. [ST:XCL] -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: z990 multiprocessor overhead issue.
On Fri, 2006-03-24 at 18:48 -0500, Richards.Bob wrote: With the CPs being pooled, I doubt the specialties draw down the MIP MP effect. They might chew some quick cycles, but they should not impact MIP capacity like additional general purpose CPs would. My take on this is that they are CPs available for (MVS) dispatch. So they see all the needed storage - system control blocks, JVM, whatever. That spells serialization requirements - and that means what is colloquially called MP factor. There are no doubt fast path mods for these guys, but I'd like to see some numbers from Walt or maybe Peter Relsons group. Obviously I can't lay my hands on such a system or I would have run some numbers myself. Shane ... -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html