Ted, I'm with you on SU, mostly.
I like percentage, it's what you see in SDSF and is easier to describe back to the user (and their demanding managers). I find SU useful, but constricting, particularly because our sysplexes invariably consist of different models of machines and trying to "limit application X to 1% of the CPU" becomes an exercise in frustration when the application is required to run on at least one image on nearly every CEC at the same time. I think SU would be more useful if you could specify it by system/image, for a single application with the same name. Say you have an application named HOG on 6 systems, and you want to limit it to 1% of the system at max on a per-system basis, and you know that it is the only workload running in the SYSHOG started task service class. It would be nice if I could limit it then, on SYSA to 100 MSU, on SYSB to 300 MSU, on SYSC to 180 MSU, SYSD to 40 MSU, SYSE to 65 MSU, and SYSF to 420 MSU. But instead, I have to either make six service classes and six resource groups (what a waste of CPU to have to traverse all that), or simply allow that any one of them might get as much as 6% of the largest box it runs on (unacceptable). SU is also painful in Batch and TSO periods. I want first period TSO to last a quarter second. Real time. Not SUs. And I want my WLM policy to apply equally to all systems (across more than 12 CECs) and all images. And I want my low batch to sit in first period for 30 minutes before dropping lower. The same way on all systems. SU doesn't seem to allow that either. It becomes an approximation, that we constantly have to review and update to keep it accurate. Percentages are better, in disparate sysplexes, if you ask me, but SU still has it's (very limited) uses in an overarching standardized policy (which is much easier to maintain than having multiple policies for each plex). Still, I wish the new percentage system could allow different percentages for different images in the same resource group. Ah well. At least having percentages is a step forward, and a welcome one at that. Do you know if there has been some update to WLM to address image-level resource groups that I haven't seen in the books? Cheers, Gary Diehl MVS Support "The glass is neither half full or half empty; the engineer who designed the glass simply allowed for a 100% increase in fluid storage." -----Original Message----- From: IBM Mainframe Discussion List [mailto:[EMAIL PROTECTED] On Behalf Of Ted MacNEIL Sent: Tuesday, December 18, 2007 3:08 PM To: [email protected] Subject: Re: WLM question. >CPU percentages are a much better methodology. They adjust with processor changes. Very cool. I disagree for two reasons: 1. Some shops wish to ensure the same service for a test workload, regardless of how many times the processor is upgraded. 2. What does a percentage mean when there are multiple machines in a SYSPLEX with varying capacities? That's why I prefer SU based resource classes, which do not (necessarily) have to be changed when a processor is upgraded. ---------------------------------------------------------------------- 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
