Max,
I thought a lot about this one. My belief is that WLM does almost
nothing on a lightly loaded machine. Unless you performed this
experiment many times under rigidly controlled conditions, I would
hesitate to draw any conclusions. You did not state whether the elapsed
run time changed. I would be surprised if it had.
Enjoy your spare time.
Cheers,
Steve
[EMAIL PROTECTED] wrote:
Hi Steve and thank you a lot for replies.
<snip>
I've another question:
I saw in the past a strange thing about jobs running in a machine with a
low CPU usage, say 2% before submitting a job. We used for this job a
service class with a velocity goal of, just to say, 30. When running this
job used for instance 10% of cpu. We noticed that modifying service class
using another with a more agressive velocity goal (say 50) it used by far
more CPU, say 20%.
What I want to say is it seems that when a job is running in a machine
with a lot of available CPU it uses the only CPU needed to reach the given
goal even if it could use potentially ALL CPU. Using a service class
more agressive causes a raise in cpu usage, but only for reaching the
higher goal. This is in contrast with some answers I received in the past
ie that this job should use all CPU available.
Any hint/tip ?
Thank you a lot and best regards
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
