Natassa, You can put an automation to RESET SRVCLASS of some none critical heavy CPU batch jobs that steal your critical MSUs. In our case we have a matrix of some non critical jobs that spend very much CPU running during morning 7:00-9:00). We have implemented an automation to RESET SRVCLASS of these "bad" jobs to a very low importance (only some SUs) in case on of these jobs not end at a specific time (8:00 am). So we prohibit the impact on our rolling 4h MSU average at the time of 12:00.
Kind regards Kostas On Thu, Feb 21, 2013 at 3:09 PM, Natasa Savinc < [email protected]> wrote: > Thank you all for comments. > Joel, this is exactly what happens: if our rolling 4-hour MSU average > starts with too high numbers in the morning, we are in trouble at noon. Of > course that we monitor it and manage it, and constantly working on > prevention (Werner, thank you for the idea of assigning classes dynamically > using REXX - we'll see if it's suitable for us) . > Still, sometimes capping happens. Then we have what Adam said, the > situation that everything is either at the same priority or importance > levels, competing for CPU cycles. During the same period of time that high > DP workload is stalling and experiencing performance degradation, batch > jobs with lowest dispatching priority manage to 'steal' certain amount of > CPU. Therefore question if not allowing those jobs to execute would ensure > that those CPU cycles are spent on high DP workload. > > Natasa > > ---------------------------------------------------------------------- > For IBM-MAIN subscribe / signoff / archive access instructions, > send email to [email protected] with the message: INFO IBM-MAIN > ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO IBM-MAIN
