Re: HiperDispatch with mostly small LPARs
I think it will depend on how many books you have. HD will likely help your largest LPAR because HD will give exclusive access to 8 physical CPs across 2 chips which will result in better cache utilization. We have 2 LPARS across 8 Physicals with 45%/55% weight distribution and are seeing about 5% benefit from HD. Turn on SMF113s before you turn on HD so that you can measure the uplift. I just gave a pitch at SHARE and you can see some of SMF 113 metrics: http://share.confex.com/share/117/webprogram/Session9689.html Of course it also depends on your workload. Regards, Santosh -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@bama.ua.edu with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: HiperDispatch with mostly small LPARs
The OLD ROT for LP:CP was 2:1. More recent hardware (e.g. 2817) have increased this limit to about 3:1 The exact value between 2:1 and 3:1 depends on the level of pain you can endure. YMMV. The amount of overhead due to context switching when another LP is dispatched varies with the activity and quantity of the LPs. DANGER! Will Robinson DANGER! This IS NOT a linear equation. IMO it tends towards cubic once the 3:1 ratio is exceeded! HTH, snip We are thinking about HiperDispatch implementation but I am not sure it is appropriate in our environment. We have a z196 Host = 2817-M66/700 with 12 CPs. There is one large LPAR with a share that allows access to 8.89 CPs. It is assigned 12 LPs. The rest of the LPARs on this box have very low shares. There are 8 additional LPARs and the highest physical processor share is 1.05. The rest of the PP shares vary from 0.16 to 0.53. I am not sure that there is any point to turning on Hiperdispatch for these LPARs. I read an article by Don Deese where he said HiperDispatch Management Mode cannot be effective in an LPAR unless the LPAR has a share of CPC capacity that results in at least 0.5 (the #CP value) equivalent physical processors. This was in a document about the z10 and I am wondering whether it is also true on a z196. If we enable HiperDispatch on the one LPAR will there be a negative effect on the other LPARs? In total now there are 37 LPs assigned for the box. I think this is a high number. /snip -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@bama.ua.edu with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: HiperDispatch isn't working on a z10
Have you done anything with your WLM Service Policy to support HD? Do you have much Web work? If you use HD, I would not suggest playing manual IRD. Configure each LPAR with the number of engines need to handle the peak work; the HD will park the excess as needed. zNorman -Original Message- From: IBM Mainframe Discussion List [mailto:ibm-m...@bama.ua.edu] On Behalf Of Petersen, Jim Sent: Wednesday, July 14, 2010 Wednesday 11:51 AM To: IBM-MAIN@bama.ua.edu Subject: HiperDispatch isn't working on a z10 Here is a problem we opened with IBM and are wondering if anyone else has observed this as well. We have currently turned off HiperDispatch because IBM wants the CE to gather data and because of other issues, we are very uncomfortable with that. We're running z/OS 1.10 in a parallell sysplex. One of our CEC's is a 2097-717 with 4 LPARs defined. One of the LPARs is not active. The weights of the LPARs is as follows: LPAR1 - 66% LPAR2 - 20% LPAR3 - 10% The LPAR that is down is LPAR4 and is weighted at 4%. During the online day, our CPU is busy but at night there is excess capacity. The LPAR LPAR3 during the day has two CPs configured online. At 19:00 when the demand drops, we configure four more engines online. As the applications throw more work to LPAR3, the utilization gets to about 11.5% but that's it. It won't take any more even though the entire CEC is about 70-80% busy. I thought that as long as the CEC had capacity, the LPAR could take more than it's weight. When I look at the engines through our monitor, the four engines that I configured online are in a PARKED state. Is this working correctly or is there something else I should do to allow more work to flow to LPAR3 during off peak times? ___ Jim Petersen MVS - Lead Systems Engineer Home Depot Technology Center 1300 Park Center Drive, Austin, TX 78753 www.homedepot.com email:jim_peter...@homedepot.com 512-977-2615 direct 512-977-2930 fax 210-859-9887 cell phone The information in this Internet Email is confidential and may be legally privileged. It is intended solely for the addressee. Access to this Email by anyone else is unauthorized. If you are not the intended recipient, any disclosure, copying, distribution or any action taken or omitted to be taken in reliance on it, is prohibited and may be unlawful. When addressed to our clients any opinions or advice contained in this Email are subject to the terms and conditions expressed in any applicable governing The Home Depot terms of business or client engagement letter. The Home Depot disclaims all responsibility and liability for the accuracy and content of this attachment and for any damages or losses arising from any inaccuracies, errors, viruses, e.g., worms, trojan horses, etc., or other items of a destructive nature, which may be contained in this attachment and shall not be liable for direct, indirect, consequential or special damages in connection with this e-mail message or its attachment. - The information contained in this e-mail and any attached 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 sent 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. -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@bama.ua.edu with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@bama.ua.edu with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: HIPERDISPATCH (was Re: First z10 BC on West Coast)
On Mon, 8 Dec 2008 16:46:15 -0600, Mark Zelden [EMAIL PROTECTED] wrote: 2) The majority of installations that bring in a z10 are not turning on I did not know that. Where did you hear that? Cheryl's newsletter? Another source? The LSPR numbers are with HIPERDISPATCH=ON. We already have trouble getting out of the box what zPCR tells us to expect (no surprise, we've had that problem since z990). There is no way we could have put one of these puppies in without starting off with HIPERDISPATCH=ON unless we wanted to purchase some (very expensive!) extra engines.I suspect the same is true with most if not all medium to large shops. We weren't the first kids on our block to get a upgrade this time. I wonder if IBM let customers turn on extra engines for free while some of the bugs with HIPERDISPATCH were being worked out (still ISV issues with this approach). Of course I meant HIPERDISPATCH=YES as far as the correct setting in IEAOPTxx, but you get the idea... Mark -- Mark Zelden Sr. Software and Systems Architect - z/OS Team Lead Zurich North America / Farmers Insurance Group - ZFUS G-ITO mailto:[EMAIL PROTECTED] z/OS Systems Programming expert at http://expertanswercenter.techtarget.com/ Mark's MVS Utilities: http://home.flash.net/~mzelden/mvsutil.html -- 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: Hiperdispatch
At the current moment, IHASVT bit SVTAFFON does not necessarily match the IEAOPTxx Hiperdispatch setting, but does reflect whether at the current time the system is in hiperdispatch mode (there are transition states where the two might not match). We have no plans to change, but conversely no intention of committing to keep, this behavior. Peter Relson z/OS Core Technology Design -- 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: Hiperdispatch
On Thu, 30 Oct 2008 07:19:32 -0400, Peter Relson [EMAIL PROTECTED] wrote: At the current moment, IHASVT bit SVTAFFON does not necessarily match the IEAOPTxx Hiperdispatch setting, but does reflect whether at the current time the system is in hiperdispatch mode (there are transition states where the two might not match). They will never match on a pre-z10 machine if you specify HIPERDISPATCH=YES in IEAOPTxx. But the WLMOPT tool which looks at the IEAOPTxx setting and SVTAFFON does report both values correctly AFAIK. Mark -- Mark Zelden Sr. Software and Systems Architect - z/OS Team Lead Zurich North America / Farmers Insurance Group - ZFUS G-ITO mailto:[EMAIL PROTECTED] z/OS Systems Programming expert at http://expertanswercenter.techtarget.com/ Mark's MVS Utilities: http://home.flash.net/~mzelden/mvsutil.html -- 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: Hiperdispatch
For me the runtime setting is fine. I believe the OPT value comes from oco cblock IRARMCTY. Thanks Peter Roland On Thu, 30 Oct 2008 07:19:32 -0400, Peter Relson [EMAIL PROTECTED] wrote: At the current moment, IHASVT bit SVTAFFON does not necessarily match the IEAOPTxx Hiperdispatch setting, but does reflect whether at the current time the system is in hiperdispatch mode (there are transition states where the two might not match). They will never match on a pre-z10 machine if you specify HIPERDISPATCH=YES in IEAOPTxx. But the WLMOPT tool which looks at the IEAOPTxx setting and SVTAFFON does report both values correctly AFAIK. -- 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: Hiperdispatch
Roland Schiradin wrote: George, according to the macro changes this flag was introduced around 2005 $LN=AFFDISP HBB7730 050311 PD00GD: Affinity Dispatcher Well this flag is off on my z9. You're saying the flag is OFF even when HiperDispatch is enabled? -- Edward E Jaffe Phoenix Software International, Inc 5200 W Century Blvd, Suite 800 Los Angeles, CA 90045 310-338-0400 x318 [EMAIL PROTECTED] http://www.phoenixsoftware.com/ -- 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: Hiperdispatch
Ed, no I don't have HIPERDISPATCH=YES as it makes no sense on a z9. I just wonder why this flag was added in 2005 long before Hiperdispatch comes out and if it's the correct flag. Roland Roland Schiradin wrote: George, according to the macro changes this flag was introduced around 2005 $LN=AFFDISP HBB7730 050311 PD00GD: Affinity Dispatcher Well this flag is off on my z9. You're saying the flag is OFF even when HiperDispatch is enabled? -- Edward E Jaffe -- 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: Hiperdispatch
Roland Schiradin wrote: Ed, no I don't have HIPERDISPATCH=YES as it makes no sense on a z9. snip Actually, it does not work on a z9 no matter what you specify in parmlib. A z10 is required for HiperDispatch. So, assuming this is the correct bit and that it's descriptively named (neither of which I know for sure), it should never be set on a z9. -- John Eells z/OS Technical Marketing IBM Poughkeepsie [EMAIL PROTECTED] -- 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: Hiperdispatch
So the support was there before the processor was available. You need a z10 for HIPERDISPATCH. My testing shows that it is the correct flag and it is set correctly. I guess I'll add that to IPLINFO. :-) Mark -- Mark Zelden Sr. Software and Systems Architect - z/OS Team Lead Zurich North America / Farmers Insurance Group - ZFUS G-ITO mailto:[EMAIL PROTECTED] z/OS Systems Programming expert at http://expertanswercenter.techtarget.com/ Mark's MVS Utilities: http://home.flash.net/~mzelden/mvsutil.html On Wed, 29 Oct 2008 00:39:37 -0500, Roland Schiradin [EMAIL PROTECTED] wrote: George, according to the macro changes this flag was introduced around 2005 $LN=AFFDISP HBB7730 050311 PD00GD: Affinity Dispatcher Well this flag is off on my z9. Roland Is there a cblock which contains an indicator for this? SVTAFFON X'80' Affinity dispatching is active Regards, George Kozakos z/OS Function Test/Level 3 Supervisor -- 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: Hiperdispatch
WLMOPT tools shows the IEAOPT and the runtime HIPERDISPATCH. Is this the IEAOPT setting or the runtime? Thank you for the information and verification. Yep good idea to add this to IPLINFO. Will do the same for SHOWzOS. During my research I also made some other changes MCCAFCTH in IEAOPT was wrong, added display for BLWLTRPCT and BLWLINTHD Perhaps a display for all IEAOPT settings is good idea for IPLINFO Roland So the support was there before the processor was available. You need a z10 for HIPERDISPATCH. My testing shows that it is the correct flag and it is set correctly. I guess I'll add that to IPLINFO. :-) Mark -- -- 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: Hiperdispatch
On Wed, 29 Oct 2008 15:35:58 -0500, Roland Schiradin [EMAIL PROTECTED] wrote: WLMOPT tools shows the IEAOPT and the runtime HIPERDISPATCH. Is this the IEAOPT setting or the runtime? Thank you for the information and verification. It shows both. In your case if you turned it on for a z9 it would look like this: HIPERDISPATCHYes, No Hiperdispatch value(inOPT, Running) Yep good idea to add this to IPLINFO. Already done, one my web site (with a few other updates) and sent to Sam for the CBT. Best Regards / Mit freundlichen Grüßen, Mark -- Mark Zelden Sr. Software and Systems Architect - z/OS Team Lead Zurich North America / Farmers Insurance Group - ZFUS G-ITO mailto:[EMAIL PROTECTED] z/OS Systems Programming expert at http://expertanswercenter.techtarget.com/ Mark's MVS Utilities: http://home.flash.net/~mzelden/mvsutil.html -- 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: Hiperdispatch
George, according to the macro changes this flag was introduced around 2005 $LN=AFFDISP HBB7730 050311 PD00GD: Affinity Dispatcher Well this flag is off on my z9. Roland Is there a cblock which contains an indicator for this? SVTAFFON X'80' Affinity dispatching is active Regards, George Kozakos z/OS Function Test/Level 3 Supervisor -- 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: HIPERDISPATCH=YES and z9
APAR Identifier .. OA26225 Last Changed 08/08/21 VARYCPU MANAGEMENT DISABLED WHEN RUNNING ON A Z9 AND SPECIFYING HIPERDISPATCH=YES Symptom .. IN INCORROUT Status ... INTRAN Severity ... 2 Date Closed . Component .. 5752SC1CX Duplicate of Reported Release . 730 Fixed Release Component Name 5752 SYS RSRCE Special Notice Current Target Date .. Flags SCP ... Platform Status Detail: Not Available PE PTF List: PTF List: Parent APAR: Child APAR list: ERROR DESCRIPTION: When running on a z9 and specifying in the IEAOPTxx parmlib member HIPERDISPATCH=YES and VARYCPU=YES (or the default of VARYCPU=YES is taken), vary cpu management is disabled. When running on z9 hardware Hiperdispatch is not possible, so IRABAINI will not enable Hiperdispatch, but it also does not re-enable vary cpu management. LOCAL FIX: When running on a z9, do not specifiy HIPERDISPATCH=YES in the IEAOPTxx parmlib member -- 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