Re: WLM managed workload
Ituriel, I agree that it is not WLM's goal to balance systems but to achieve goals. So if you have a 30% and 90% lpar, both running according to goal, WLM is happy. If one goal is not met, will shift workload by starting and stopping its initiators. However, this process stops far too rapidly when all systems are above 95% used, because WLM considers them both full and sees no use in more managing. It is this decision that I want to be more elaborated in/by WLM. It is perfectly possible to improve bad PIs in systems that are 98% and 100% used by shifting jobs b.m.o. starting and stopping initiators. I do this by hand and I see no reason why WLM can't do this too. Kees. -Original Message- From: IBM Mainframe Discussion List [mailto:IBM-MAIN@LISTSERV.UA.EDU] On Behalf Of ITURIEL DO NASCIMENTO NETO Sent: Wednesday, July 24, 2013 17:40 To: IBM-MAIN@LISTSERV.UA.EDU Subject: RES: WLM managed workload Kees, The intent of JES2 and WLM is not to balance batch workload. WLM does not want that jobs, transactions be fast, WLM wants them happy, and for WLM, happy means PI less than 1. We have seen in the past a terrible work done by JES2 and WLM when spreading batch jobs in several images of a JESPLEX, where a Lpar was running at 90% while others were running at 30%. Despite Load discrepancy, all goals were been met, but it's ugly to see and hard to explain. Actually there is a redbook that explains in detail how they work together, and it is very illustrative (http://www.redbooks.ibm.com/redbooks/pdfs/sg246472.pdf). I think there is no easy way to do what you want, if there is. Our goal here was to spread batch workload in a way that all Lpars should be running with similar load, and to do so, we have developed a batch load scheduling that put things in order, and now we see no big differences among Lpars involved. Atenciosamente / Regards / Saludos Ituriel do Nascimento Neto BANCO BRADESCO S.A. 4250 / DPCD Engenharia de Software Sistemas Operacionais Mainframes Tel: +55 11 3684-2177 R: 42177 3-1402 Fax: +55 11 3684-4427 Agora é BRA. BRA de Brasil. BRA de Bradesco. Patrocinador oficial dos Jogos Olímpicos e Paralímpicos Rio 2016. -Mensagem original- De: IBM Mainframe Discussion List [mailto:IBM-MAIN@LISTSERV.UA.EDU] Em nome de Vernooij, CP - SPLXM Enviada em: quarta-feira, 24 de julho de 2013 10:38 Para: IBM-MAIN@LISTSERV.UA.EDU Assunto: Re: WLM managed workload Allan, With WLM managing where work is initiated, I meant, that, when more capacity is needed, WLM starts WLM Managed Initiators where capacity is available and stops them on systems that are full. By means of this it has control over where jobs in WLM managed jobclasses are initiated and doint so it can route more work to one system and less work to another. In other words, it can help balance the load of the systems in the sysplex. My complaint/wish is, that it should do more to route work from full and overloaded systems to full but yet not overloaded systems. However, it considers systems over 95% utilized equally as full, as if they cannot be helped anymore and I think they can. Kees. -Original Message- From: IBM Mainframe Discussion List [mailto:IBM-MAIN@LISTSERV.UA.EDU] On Behalf Of Staller, Allan Sent: Wednesday, July 24, 2013 14:50 To: IBM-MAIN@LISTSERV.UA.EDU Subject: Re: WLM managed workload WLM considers all LPARs within the sysplex to be equal. WLM by itself, has no control over where the work is initiated. In conjunction with JES2 at the z/OS 2.4 level and higher, the ability was created to allow work to be started via workload managed initiators. Without going into the details, job queue time is now included as part of the WLM performance index when deciding whether to add work to the system. Also with WLM managed initiators, the current PI for the work on a given LPAR Is taken into consideration when deciding where to initiate the work (subject to other considerations(scheduling environment, system affinity,.) ). This can also be complicated by a multiple JESPLEX with a SYSPLEX. The short story on Workload Managed initiators is as follows: 1) JES2 and WLM initiators can be mixed at will. Each jobclass defined/used should be either JES or WLM managed, but not both. i.e. a given jobclass is either workload managed or jes managed. 2) A separate WLM service class should be defined for the WLM managed work to prevent misleading PI's based on the queue time component of the WLM PI calculation. There is a lot more to this, and I suggest reading the JES2 and WLM manuals extensively before proceeding. There are also several presentations from SHARE, that are quite helpful. HTH, snip WLM takes the load and performance of LPARs in consideration, by starting and stopping WLM Managed Initiators on systems that have capacity available or are overloaded. But it does this very coarse, putting a line at 95% utilization. When you have several LPARs
Re: WLM managed workload
Thanks Kees Ituriel, Still i am struggling to tell WLM how PI of a workload may have different acceptance level on different LPARs. For Me I just would like WLM to schedule workload on one LPAR unless that LPAR can't run it or is not available. I know i can do that by establishing/ controlling RESOURCE group / Scheduling environment via MVS commands but would be easier If i can define this rule within WLM policies. Thanks for your help. regards Munif. -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: WLM managed workload
Hello Munif, WLM takes the load and performance of LPARs in consideration, by starting and stopping WLM Managed Initiators on systems that have capacity available or are overloaded. But it does this very coarse, putting a line at 95% utilization. When you have several LPARs that are running near 100%, some of which are overloaded with a bad PI and others still can take more load, WLM considers them all the same. I am looking too at a mechanism to manage WLM managed batch on this type of LPARs. I think the way I go is by allowing or withholding certain jobclasses from certain systems with the $tjobclass(a,b,c,d),QAFF=-name and QAFF=+name. I do this now manually every now and then when I see an overloaded system, but I am still looking for an interface to the information you mention (PI, capping) to automate this. it is really a miss, that WLM does not do this more detailed. Kees. -Original Message- From: IBM Mainframe Discussion List [mailto:IBM-MAIN@LISTSERV.UA.EDU] On Behalf Of Munif Sadek Sent: Wednesday, July 24, 2013 07:02 To: IBM-MAIN@LISTSERV.UA.EDU Subject: WLM managed workload Dear Listers Can some one please point me in the right direction to implement WLM managed job class that can schedule jobs on different LPARS selected on basis of service class Performance Index and/or soft CAPPING status of individual LPAR. Creating Scheduling Resources / Environment and using automation to turn off/on resources is a possibility but I prefer the first option.. By the way it is DB2 9.0, z/OS 1.12 parallel sysplex environment. Regards, Munif -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN For information, services and offers, please visit our web site: http://www.klm.com. This e-mail and any attachment may contain confidential and privileged material intended for the addressee only. If you are not the addressee, you are notified that no part of the e-mail or any attachment may be disclosed, copied or distributed, and that any other action related to this e-mail or attachment is strictly prohibited, and may be unlawful. If you have received this e-mail by error, please notify the sender immediately by return e-mail, and delete this message. Koninklijke Luchtvaart Maatschappij NV (KLM), its subsidiaries and/or its employees shall not be liable for the incorrect or incomplete transmission of this e-mail or any attachments, nor responsible for any delay in receipt. Koninklijke Luchtvaart Maatschappij N.V. (also known as KLM Royal Dutch Airlines) is registered in Amstelveen, The Netherlands, with registered number 33014286 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: WLM managed workload
WLM considers all LPARs within the sysplex to be equal. WLM by itself, has no control over where the work is initiated. In conjunction with JES2 at the z/OS 2.4 level and higher, the ability was created to allow work to be started via workload managed initiators. Without going into the details, job queue time is now included as part of the WLM performance index when deciding whether to add work to the system. Also with WLM managed initiators, the current PI for the work on a given LPAR Is taken into consideration when deciding where to initiate the work (subject to other considerations(scheduling environment, system affinity,.) ). This can also be complicated by a multiple JESPLEX with a SYSPLEX. The short story on Workload Managed initiators is as follows: 1) JES2 and WLM initiators can be mixed at will. Each jobclass defined/used should be either JES or WLM managed, but not both. i.e. a given jobclass is either workload managed or jes managed. 2) A separate WLM service class should be defined for the WLM managed work to prevent misleading PI's based on the queue time component of the WLM PI calculation. There is a lot more to this, and I suggest reading the JES2 and WLM manuals extensively before proceeding. There are also several presentations from SHARE, that are quite helpful. HTH, snip WLM takes the load and performance of LPARs in consideration, by starting and stopping WLM Managed Initiators on systems that have capacity available or are overloaded. But it does this very coarse, putting a line at 95% utilization. When you have several LPARs that are running near 100%, some of which are overloaded with a bad PI and others still can take more load, WLM considers them all the same. I am looking too at a mechanism to manage WLM managed batch on this type of LPARs. I think the way I go is by allowing or withholding certain jobclasses from certain systems with the $tjobclass(a,b,c,d),QAFF=-name and QAFF=+name. I do this now manually every now and then when I see an overloaded system, but I am still looking for an interface to the information you mention (PI, capping) to automate this. it is really a miss, that WLM does not do this more detailed. Kees. -Original Message- From: IBM Mainframe Discussion List [mailto:IBM-MAIN@LISTSERV.UA.EDU] On Behalf Of Munif Sadek Sent: Wednesday, July 24, 2013 07:02 To: IBM-MAIN@LISTSERV.UA.EDU Subject: WLM managed workload Dear Listers Can some one please point me in the right direction to implement WLM managed job class that can schedule jobs on different LPARS selected on basis of service class Performance Index and/or soft CAPPING status of individual LPAR. Creating Scheduling Resources / Environment and using automation to turn off/on resources is a possibility but I prefer the first option.. By the way it is DB2 9.0, z/OS 1.12 parallel sysplex environment. /snip -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: WLM managed workload
snip With WLM managing where work is initiated, I meant, that, when more capacity is needed, WLM starts WLM Managed Initiators where capacity is available and stops them on systems that are full. By means of this it has control over where jobs in WLM managed jobclasses are initiated and doint so it can route more work to one system and less work to another. In other words, it can help balance the load of the systems in the sysplex. /snip Agreed. snip My complaint/wish is, that it should do more to route work from full and overloaded systems to full but yet not overloaded systems. However, it considers systems over 95% utilized equally as full, as if they cannot be helped anymore and I think they can. /snip This still comes down to the PI calculations for the individual workload(s). -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
RES: WLM managed workload
Kees, The intent of JES2 and WLM is not to balance batch workload. WLM does not want that jobs, transactions be fast, WLM wants them happy, and for WLM, happy means PI less than 1. We have seen in the past a terrible work done by JES2 and WLM when spreading batch jobs in several images of a JESPLEX, where a Lpar was running at 90% while others were running at 30%. Despite Load discrepancy, all goals were been met, but it's ugly to see and hard to explain. Actually there is a redbook that explains in detail how they work together, and it is very illustrative (http://www.redbooks.ibm.com/redbooks/pdfs/sg246472.pdf). I think there is no easy way to do what you want, if there is. Our goal here was to spread batch workload in a way that all Lpars should be running with similar load, and to do so, we have developed a batch load scheduling that put things in order, and now we see no big differences among Lpars involved. Atenciosamente / Regards / Saludos Ituriel do Nascimento Neto BANCO BRADESCO S.A. 4250 / DPCD Engenharia de Software Sistemas Operacionais Mainframes Tel: +55 11 3684-2177 R: 42177 3-1402 Fax: +55 11 3684-4427 Agora é BRA. BRA de Brasil. BRA de Bradesco. Patrocinador oficial dos Jogos Olímpicos e Paralímpicos Rio 2016. -Mensagem original- De: IBM Mainframe Discussion List [mailto:IBM-MAIN@LISTSERV.UA.EDU] Em nome de Vernooij, CP - SPLXM Enviada em: quarta-feira, 24 de julho de 2013 10:38 Para: IBM-MAIN@LISTSERV.UA.EDU Assunto: Re: WLM managed workload Allan, With WLM managing where work is initiated, I meant, that, when more capacity is needed, WLM starts WLM Managed Initiators where capacity is available and stops them on systems that are full. By means of this it has control over where jobs in WLM managed jobclasses are initiated and doint so it can route more work to one system and less work to another. In other words, it can help balance the load of the systems in the sysplex. My complaint/wish is, that it should do more to route work from full and overloaded systems to full but yet not overloaded systems. However, it considers systems over 95% utilized equally as full, as if they cannot be helped anymore and I think they can. Kees. -Original Message- From: IBM Mainframe Discussion List [mailto:IBM-MAIN@LISTSERV.UA.EDU] On Behalf Of Staller, Allan Sent: Wednesday, July 24, 2013 14:50 To: IBM-MAIN@LISTSERV.UA.EDU Subject: Re: WLM managed workload WLM considers all LPARs within the sysplex to be equal. WLM by itself, has no control over where the work is initiated. In conjunction with JES2 at the z/OS 2.4 level and higher, the ability was created to allow work to be started via workload managed initiators. Without going into the details, job queue time is now included as part of the WLM performance index when deciding whether to add work to the system. Also with WLM managed initiators, the current PI for the work on a given LPAR Is taken into consideration when deciding where to initiate the work (subject to other considerations(scheduling environment, system affinity,.) ). This can also be complicated by a multiple JESPLEX with a SYSPLEX. The short story on Workload Managed initiators is as follows: 1) JES2 and WLM initiators can be mixed at will. Each jobclass defined/used should be either JES or WLM managed, but not both. i.e. a given jobclass is either workload managed or jes managed. 2) A separate WLM service class should be defined for the WLM managed work to prevent misleading PI's based on the queue time component of the WLM PI calculation. There is a lot more to this, and I suggest reading the JES2 and WLM manuals extensively before proceeding. There are also several presentations from SHARE, that are quite helpful. HTH, snip WLM takes the load and performance of LPARs in consideration, by starting and stopping WLM Managed Initiators on systems that have capacity available or are overloaded. But it does this very coarse, putting a line at 95% utilization. When you have several LPARs that are running near 100%, some of which are overloaded with a bad PI and others still can take more load, WLM considers them all the same. I am looking too at a mechanism to manage WLM managed batch on this type of LPARs. I think the way I go is by allowing or withholding certain jobclasses from certain systems with the $tjobclass(a,b,c,d),QAFF=-name and QAFF=+name. I do this now manually every now and then when I see an overloaded system, but I am still looking for an interface to the information you mention (PI, capping) to automate this. it is really a miss, that WLM does not do this more detailed. Kees. -Original Message- From: IBM Mainframe Discussion List [mailto:IBM-MAIN@LISTSERV.UA.EDU] On Behalf Of Munif Sadek Sent: Wednesday, July 24, 2013 07:02 To: IBM-MAIN@LISTSERV.UA.EDU Subject: WLM managed workload Dear Listers Can some one please point me in the right direction to implement WLM managed job
WLM managed workload
Dear Listers Can some one please point me in the right direction to implement WLM managed job class that can schedule jobs on different LPARS selected on basis of service class Performance Index and/or soft CAPPING status of individual LPAR. Creating Scheduling Resources / Environment and using automation to turn off/on resources is a possibility but I prefer the first option.. By the way it is DB2 9.0, z/OS 1.12 parallel sysplex environment. Regards, Munif -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
