Menu
▾
▴
Re: [Lse-tech] idle thread's cpus_allowed
|
From: Mala A. <ma...@us...> - 2002-12-16 17:55:19
|
John-
I am looking at a scheduler performance degradation problem. I see
20% performance
degradation using O(1) scheduler (RH AS) v/s the old scheduler (Red Hat
7.2).
I see the O(1) in 2.5 kernels have some changes. Is there a patch
available that
back ports 2.5 changes to 2.4 kernel?
I see that the number of processes in the running state goes down using
O(1)
scheduler (RH AS) from 15 to 8. I have not pinned down the processes to
processors.
Are you planning on fixing the load_balance problem? Please let me know.
Thanks.
Regards,
Mala
Mala Anand
IBM Linux Technology Center - Kernel Performance
E-mail:ma...@us...
http://www-124.ibm.com/developerworks/opensource/linuxperf
Phone:838-8088; Tie-line:678-8088
"John Hawkes"
<ha...@sg...> To: kr...@us...
Sent by: cc: "lse-tech" <lse...@li...>, "David Mosberger"
lse...@li...ur <da...@hp...>
ceforge.net Subject: Re: [Lse-tech] idle thread's cpus_allowed
12/13/02 01:48 PM
From: "Mike Kravetz" <kr...@us...>
> Are you actually seeing idle threads migrate, or is your concern
> based on the uninitialized cpus_allowed fields?
We're actually seeing idle threads migrate.
You're not going to see this without having that "busiest" CPU filled
with pinned-down processes.
Which brings up another flaw in the load_balance() algorithm that
eventually needs to be addressed: If the "busiest" CPU has no
migratable processes, then load_balance() just gives up, rather than
continuing to search on the next-busiest CPU's runqueue.
John Hawkes
-------------------------------------------------------
This sf.net email is sponsored by:
With Great Power, Comes Great Responsibility
Learn to use your power at OSDN's High Performance Computing Channel
http://hpc.devchannel.org/
_______________________________________________
Lse-tech mailing list
Lse...@li...
https://lists.sourceforge.net/lists/listinfo/lse-tech
|