Just as a note, on V1R9, and rolled back to V1R7 and V1R8, there is now Blocked Workload/trickle support, which, if configured, will make sure all work, even discretionary, gets some access to the CPU. You can configure how much of the CPU you want to give discretionary work, assuming your CPU is 100% utilized. This might forestall screams. And if you're willing to enable WLM-managed initiators, this can ameliorate the problem of a job tying up an initiator, since WLM will start more if it decides there is capacity available.
See http://www-03.ibm.com/support/techdocs/atsmastr.nsf/WebIndex/FLASH10609 for more information. --- Kevin McKenzie External Phone: 845-435-8282, Tie-line: 8-295-8282 z/OS BCP SVT, Dept FXKA, Bldg 706/2D38 From: Ed Gould <[email protected]> To: [email protected] Date: 07/29/2009 05:40 PM Subject: Re: Enforcing CPU Time Sent by: IBM Mainframe Discussion List <[email protected]> --- On Wed, 7/29/09, Ted MacNEIL <[email protected]> wrote: From: Ted MacNEIL <[email protected]> Subject: Re: Enforcing CPU Time To: [email protected] Date: Wednesday, July 29, 2009, 2:13 PM >When you've set a cap that you constantly hit, then you need to control that which is causing you to hit the cap, especially when it is a case of a test system causing problems with a production system. A cap is better than cancelling. I'd rather have a test job hang than blow it away and come back and use more resources later. That's what 'Discretionary' is for. Ted: I am mixed on your answer. The problem as I see it if you cap it it ties up the initiator for a potentially long time. It also could mean that you do not meet service levels and your management gets dinged (and the politics that follows). Yet if you"cancel" the job (322 or 222 or whatever) you can free up the initiator for other test jobs to execute. Now days the operator does NOT sit around and monitor jobs so there is no easy way for him/her to know that the job is essentially sitting there. In a busy shop I could see a job capped and no thruput occurs and then the screams start I CAN't get my job done etc etc usually its late in the day before anyone gets upset and the programmers get hammered as they are doing "nothing". On the overall scheme of things people can live with 322 or 122 or whatever a lot easier than no job thruput. A case in point we had a shared spool and 3 test job classes were set up so test could run with IDMS (no idea what type of jobs other than they needed IDMS). If those test jobclass got delayed the phones started to ring off the hook demanding to know "WHY"!!! . The answer is NOT to fire off another initiator as we had it finely tuned and (IIRC) only one job could be running against IDMS at the time. Hey I didn't design it we just implemented what the IDMS person wanted. There were also other issues of resource availability (tape drives and the like). This was a real world situation. We even went to the programmers management and explained the situation and the agreed that a 322 or 522 or what ever was acceptable rather than delaying other people. Ed ---------------------------------------------------------------------- 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
