Max,
I know it might just be a bit of confusion, but there is no priority
parameter in WLM. You might be equating importance to priority, which
leads to all kind of errors in setting up a service policy.
Something with a velocity goal of 90 and an importance of 1 should
probably be in SYSSTC, avoiding the overhead of measurement.
Your interactive work should have response time goals if supported by
the subsystem. The goals should reflect what the users expect to (want
to?) see.
Please get back to me if you have any further questions.
Best,
Steve Samson
Max Scarpa wrote:
Hi all
I completely agree that if we haven't a SLA we haven't any baseline to
check what's happening and if this is correct. What I see, other than PI >
1 for important STCs (ranging from 1.7 to 2.5 and more in some cases) is
that we have some velocity goals of 90 and priority 1 which will never be
reached in any way (max is 75-78), too aggressive goals for other service
classes and most of workload in a narrow range of priority with a lot of
service classes. It's what many experts say it's to be avoided, having a
lot of work at the 'upper limits' of WLM. Moreover all is worsened by WLC.
I asked in a meeting what happens if WLM never reach the goal of a service
class. The answer: was after a while it starts to left behind the work not
reaching the goal. But I see that it continue to maintain its DP (even if I
didn't monitor it very frequently) and it seems strange to me, but I know
that the reply came from an expert and I do not doubt about it.
And yes I'm receiving many complaints from clients but it seems to have no
effect to some other people.
Any hint about papers is welcome.
Max Scarpa
----------------------------------------------------------------------
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
