Poor performance with very high runqueue

From: Green, Simon (SGreen@KRAFTEUROPE.COM)
Date: Mon Nov 11 2002 - 12:43:52 EST


I'm working away from the office, which means I don't have all of the books
and stuff I'd normally rely on, so I hope you'll forgive me asking a few
possibly quite dumb questions.

The system I'm looking at is a 2-way SP2 Winterhawk II node. It's running
at 100% cpu utilisation, with a runqueue averaging around 50. (Low is 40;
high is 75) No evidence of memory problems.

Now the obvious problem is lack of cpus, (not simply cpu power), so
upgrading to a 4-way is attractive, but I want to get at least some idea of
whether that will help before advising someone to spend a lot of money.

I believe that the problem is caused, not by a heavy workload, but by
relatively idle applications continually polling, (every 100 ms) to see if
there's anything for them to do.

First of all, can somebody remind me of what is a reasonable runqueue size
for a given number of processors, (2 and 4, obviously). (And I know that 50
is much too high.)

Second, and more interestingly, I have another similar system which is a
4-way Winterhawk II. It's busy, but has 20% idle and a runqueue averaging
about 6; the highest I've seen is 14, so that looks pretty healthy. I would
like to disable two of the processors and see if its runqueue size suddenly
increases. How can I do this, and how do I put the processors back
afterwards?

Both systems are running AIX 4.3.3 ML10.

Thanks in advance for your help.

Simon Green
Philip Morris ITSC Europe

AIX-L Archive at http://marc.theaimsgroup.com/?l=aix-l&r=1&w=2
AIX FAQ at http://www.faqs.org/faqs/aix-faq/

N.B. Unsolicited email from vendors will seldom be appreciated.



This archive was generated by hypermail 2.1.7 : Wed Apr 09 2008 - 22:16:19 EDT