Re: Setting up PR/SM for very small Capacity LPARs on Fast Processors
On Wed, 24 Sep 2008 14:55:21 +0200, R.S. [EMAIL PROTECTED] wrote: What does it mean counterproductive ? Assuming you really need the system and the system need in MSU is really small you can do the following: 1. Assign small percentage (I've seen even sub-percent weights) despite it is counterproductive 2. Assign more power than it needs. Of course unused cycles will still be used by other LPARs if needed. 3. Look for another (smaller) CPC. Did I miss something? Perhaps, because your suggestions are not really keeping the utilization of these under - say 3%, per the OP, which I take to mean cap. 1. Unless you refrain from capping, this has been very counter-productive at our shop. WLM barely has time to figure out who to dispatch, and the LPAR is pre-empted. If you do not cap, you're really not keeping the utilization below 3% if there are cycles to spare. 2. I agree this is a valid alterntive to the what the OP is considering, but then you're not keeping the utilization under 3%. 3. Mark also suggested the penalty box alternative - fine, if you can afford it, but then you don't necessarily have to keep utilization below 3%. IBM has been after us for years (rightly so) to consolidate our tiny LPARs, where we can. The z/VM comment was a little tongue-in-cheek, but not completely out of the realm of alternatives, is it? We've opted to move workload and eliminate LPARs altogether. Regards, Art Gutowski Ford Motor Company ITI -- 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: Setting up PR/SM for very small Capacity LPARs on Fast Processors
On Tue, 23 Sep 2008 15:34:35 -0400, Knutson, Sam [EMAIL PROTECTED] wrote: We have to figure out what we will be doing as we replace our smallest CEC with a z10 for the smallest LPARs as well. z/VM? Seriously, we'll be having similar discussions, here. We have several tiny JES2 systems running alongside some comparatively huge JES3 systems and a habit of keeping LPARs to an established MIPS (or MSU) rating, though we've been advised keeping an LPAR under 3-5% of the box can be counterproductive. I would be interested in talking to you offline on this - our z10 was going in September, then October, now likely November. Regards, Art Gutowski Ford Motor Company ITI -- 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: Setting up PR/SM for very small Capacity LPARs on Fast Processors
Arthur Gutowski wrote: [...] Seriously, we'll be having similar discussions, here. We have several tiny JES2 systems running alongside some comparatively huge JES3 systems and a habit of keeping LPARs to an established MIPS (or MSU) rating, though we've been advised keeping an LPAR under 3-5% of the box can be counterproductive. What does it mean counterproductive ? Assuming you really need the system and the system need in MSU is really small you can do the following: 1. Assign small percentage (I've seen even sub-percent weights) despite it is counterproductive 2. Assign more power than it needs. Of course unused cycles will still be used by other LPARs if needed. 3. Look for another (smaller) CPC. Did I miss something? Regards -- Radoslaw Skorupka Lodz, Poland -- BRE Bank SA ul. Senatorska 18 00-950 Warszawa www.brebank.pl Sd Rejonowy dla m. st. Warszawy XII Wydzia Gospodarczy Krajowego Rejestru Sdowego, nr rejestru przedsibiorców KRS 025237 NIP: 526-021-50-88 Wedug stanu na dzie 01.01.2008 r. kapita zakadowy BRE Banku SA wynosi 118.642.672 zote i zosta w caoci wpacony. -- 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: Setting up PR/SM for very small Capacity LPARs on Fast Processors
On Wed, 24 Sep 2008 07:46:56 -0500, Arthur Gutowski [EMAIL PROTECTED] wrote: On Tue, 23 Sep 2008 15:34:35 -0400, Knutson, Sam [EMAIL PROTECTED] wrote: We have to figure out what we will be doing as we replace our smallest CEC with a z10 for the smallest LPARs as well. z/VM? Seriously, we'll be having similar discussions, here. We have several tiny JES2 systems running alongside some comparatively huge JES3 systems and a habit of keeping LPARs to an established MIPS (or MSU) rating, though we've been advised keeping an LPAR under 3-5% of the box can be counterproductive. I would be interested in talking to you offline on this - our z10 was going in September, then October, now likely November. With the scalability of the various z10 (and z9) models, I don't understand why this is an issue. Aren't the 4xx models an option? BTW, we've had a 2 engine penalty box that was a z900, then z9 and soon to be z10. The 3 smallest LPARs (plus a sandbox LPAR) have always run with a single CP and we've never had a problem on those LPARs due to the single engine. As I mentioned in an earlier post, we've had 4:1 (or higher) LP to CP ratios on that box. 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
Setting up PR/SM for very small Capacity LPARs on Fast Processors
Hi. I'm working with migrating a customer to z9 hardware, and when it comes to the PR/SM setup I have a few questions, which pops up from time to time I believe. When it comes to assignning Logical Processor to an LPAR, do we still have a ROT of 2:1 ratio between Logical and Physical Processors on the CEC, - or has the PR/SM overhead been minimized over time. Another thing : Way back we didn't assign just 1 LP to an LPAR regardless of how small the LPAR was, - to come around the problems arising with looping applications on an LPAR with just 1 LP. But we are discussing whether or not it's appropriate with z9 to assign 2 LP to a small Capacity LPAR, and keep the utilization of these under - say 3%. Any comments on these issues would be appriciated. Thanks in advance Med venlig hilsen / Kind regards Flemming Sigvardt -- IBM, IT Delivery, Denmark Borupvang 1, DK-2750 Ballerup Office : (+45) 4523 8467 / Cell : (+45) 2880 8467 Email : [EMAIL PROTECTED] -- Medmindre andet er angivet ovenfor: / Unless Otherwise Stated Above: IBM Danmark A/S Nymøllevej 91 2800 Kongens Lyngby, Danmark CVR nr.: 65305216 -- 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: Setting up PR/SM for very small Capacity LPARs on Fast Processors
On Tue, 23 Sep 2008 12:22:27 +0200, Flemming Sigvardt [EMAIL PROTECTED] wrote: Hi. I'm working with migrating a customer to z9 hardware, and when it comes to the PR/SM setup I have a few questions, which pops up from time to time I believe. When it comes to assignning Logical Processor to an LPAR, do we still have a ROT of 2:1 ratio between Logical and Physical Processors on the CEC, - or has the PR/SM overhead been minimized over time. Another thing : Way back we didn't assign just 1 LP to an LPAR regardless of how small the LPAR was, - to come around the problems arising with looping applications on an LPAR with just 1 LP. But we are discussing whether or not it's appropriate with z9 to assign 2 LP to a small Capacity LPAR, and keep the utilization of these under - say 3%. Any comments on these issues would be appriciated. Thanks in advance There are plenty of comments and advise in the archives. Here is a post from mine on the subject from about a year ago - Oct. 2007: Subject: Logical to Physical CPU ratio ROT http://bama.ua.edu/cgi-bin/wa?A2=ind0710L=ibm-mainD=1amp;O=DP=75308 -- 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: Setting up PR/SM for very small Capacity LPARs on Fast Processors
Useful references IBM White paper WP100925, Managing CPU-Intensive Work on Uniprocessor LPARs by Kathy Walsh 2006 http://www-03.ibm.com/support/techdocs/atsmastr.nsf/WebIndex/WP100925 Managing Workloads on a Uniprocessor by Linda August 2007 http://regions.cmg.org/regions/ncacmg/downloads/mar082007_session2.pdf or http://www.share.org/EventDocuments/tabid/323/Default.aspx?document=proceedings/SHARE_in_Tampa_Bay/S2524LA084813.pdf Today we also don't let LPARs run with one processor (VARYCPUMIN= in IEAOPTxx where we have IRD managing engines). We have to figure out what we will be doing as we replace our smallest CEC with a z10 for the smallest LPARs as well. Thanks, Sam Knutson (GEICO) -Original Message- From: IBM Mainframe Discussion List [mailto:[EMAIL PROTECTED] On Behalf Of Flemming Sigvardt Sent: Tuesday, September 23, 2008 6:22 AM To: IBM-MAIN@BAMA.UA.EDU Subject: Setting up PR/SM for very small Capacity LPARs on Fast Processors Hi. I'm working with migrating a customer to z9 hardware, and when it comes to the PR/SM setup I have a few questions, which pops up from time to time I believe. When it comes to assignning Logical Processor to an LPAR, do we still have a ROT of 2:1 ratio between Logical and Physical Processors on the CEC, - or has the PR/SM overhead been minimized over time. Another thing : Way back we didn't assign just 1 LP to an LPAR regardless of how small the LPAR was, - to come around the problems arising with looping applications on an LPAR with just 1 LP. But we are discussing whether or not it's appropriate with z9 to assign 2 LP to a small Capacity LPAR, and keep the utilization of these under - say 3%. Any comments on these issues would be appriciated. Thanks in advance Med venlig hilsen / Kind regards Flemming Sigvardt -- IBM, IT Delivery, Denmark Borupvang 1, DK-2750 Ballerup Office : (+45) 4523 8467 / Cell : (+45) 2880 8467 Email : [EMAIL PROTECTED] -- Medmindre andet er angivet ovenfor: / Unless Otherwise Stated Above: IBM Danmark A/S Nymøllevej 91 2800 Kongens Lyngby, Danmark CVR nr.: 65305216 -- This email/fax message is for the sole use of the intended recipient(s) and may contain confidential and privileged information. Any unauthorized review, use, disclosure or distribution of this email/fax is prohibited. If you are not the intended recipient, please destroy all paper and electronic copies of the original message. -- 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: Setting up PR/SM for very small Capacity LPARs on Fast Processors
On Tue, 23 Sep 2008 15:34:35 -0400, Knutson, Sam [EMAIL PROTECTED] wrote: Managing Workloads on a Uniprocessor by Linda August 2007 http://regions.cmg.org/regions/ncacmg/downloads/mar082007_session2.pdf Gee... wasn't reduced preemption with MVS/ESA V5.1 (not MVS/SP 3.1)? 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: Setting up PR/SM for very small Capacity LPARs on Fast Processors
Reduced preemption was introduced with SP3.1. I don't think that I kept the microfiche for SP3.1, but I still have the microfiche for SP4.2 and Reduced Preemption was alive and well in SP4.2. Possibly you are confusing reduced preemption with the dispatcher redesign that was available with SP5.1 (and is partially described on Linda's next foil). Don ** Don Deese, Computer Management Sciences, Inc. Voice: (804) 776-7109 Fax: (8043) 776-7139 http://www.cpexpert.org ** At 03:54 PM 9/23/2008, you wrote: On Tue, 23 Sep 2008 15:34:35 -0400, Knutson, Sam [EMAIL PROTECTED] wrote: Managing Workloads on a Uniprocessor by Linda August 2007 http://regions.cmg.org/regions/ncacmg/downloads/mar082007_session2.pdf Gee... wasn't reduced preemption with MVS/ESA V5.1 (not MVS/SP 3.1)? 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 -- 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: Setting up PR/SM for very small Capacity LPARs on Fast Processors
Gee... wasn't reduced preemption with MVS/ESA V5.1 (not MVS/SP 3.1)? Yes. Unfortunately, it removed MTTW (except for DISCRETIONARY), so where I/O intensive workloads used to get in, when needed, since then they have to wait on CPU intensive ones. This wrecked havoc on our batch service when we went to goal mode in 1999. And, it can still cause problems! - Too busy driving to stop for gas! -- 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: Setting up PR/SM for very small Capacity LPARs on Fast Processors
On Tue, 23 Sep 2008 16:59:19 -0400, Don Deese [EMAIL PROTECTED] wrote: Reduced preemption was introduced with SP3.1. I don't think that I kept the microfiche for SP3.1, but I still have the microfiche for SP4.2 and Reduced Preemption was alive and well in SP4.2. Possibly you are confusing reduced preemption with the dispatcher redesign that was available with SP5.1 (and is partially described on Linda's next foil). Don Yes, you are correct.I also think I never knew about it until the SP 5 time frame when there was more information about it, WLM, the dispatcher fair share changes and preemptible SRBs (which were in SP 5.2). Thanks for the correction. 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