[GE users] Fairshare priority question: High CPU utilization followup

Stephan Grell - Sun Germany - SSG - Software Engineer stephan.grell at sun.com
Tue Mar 15 10:53:11 GMT 2005


slitster at rcn.com wrote:

>Firstly thanks to all those who repsonded so quickly to my 
>original request.
>
>I nailed the problem down to the following parameter being 
>set to low  (200 by default) -I had close to 4000 jobs queued
>
>max_functional_jobs_to_schedule 200
>
>I increased the number to 5000, and all the queued jobs 
>recieved "tickets" with appropriate scheduling.
>
>This leads to me to another question, my CPU utilization of 
>the scheduling node (Dual 3.2Xeon) is now constantly 50-55% 
>utilized (there is no iowait). Are other people seeing this 
>type of CPU utilization on a head node when they have about 
>4000 jobs queued?
>  
>
I assume, that most of the cpu time is consumed by the scheduler. To figure
to how much more expensive your setting made the functional ticket 
computation,
you can use the profiling.

How many slots does your grid have? The max_functional_jobs_to_schedule
setting does not need to be much larger than the number of slots in your 
system
to ensure a "correct" scheduling.

Cheers,
Stephan

>Cheers,
>
>Steve 
>
>---------------------------------------------------------------------
>To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
>For additional commands, e-mail: users-help at gridengine.sunsource.net
>
>  
>


---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
For additional commands, e-mail: users-help at gridengine.sunsource.net




More information about the gridengine-users mailing list