On Tue, 2013-02-19 at 04:53 -0600, Natasa Savinc wrote:
> There are two opinions amoung our sysprogs: one is that we should
> cancel all low priority workload in order to help our online get all
> the resources, the other is that that is not necessary, as batch isn't
> getting any online's CPU resources anyway.
I would agree with the latter. Assign your low priority batch as
discretionary and let the system sort it out. I once very much liked
discretionary goals and allowed most of our batch work to be processed
through it. But that was not in a CPU constrained environment.
When we implemented subcapacity pricing, we discovered the hard way that
when LPAR softcapping takes place WLM will happily underinitiate
discretionary workloads (what *you* seem to want it to do). We couldn't
have that, so now our primary batch service class looks like this:
# Duration Imp Goal description
- --------- - ----------------------------------------
1 1200 4 50% complete within 00:00:04.000
2 27000 4 95% complete within 00:01:00.000
3 4 Execution velocity of 10
... which gives us relatively good response on short jobs. Nota bene:
those duration and goal values are really installation specific. If you
do something like this, develop your own histogram from your RMF data.
--
David Andrews
A. Duda & Sons, Inc.
[email protected]
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN
