mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
* Poor scheduling when not loaded at 100% (Was: [PATCH] sched.c: Be a bit more conservative in SMP)
@ 2006-09-22 14:24 Al Boldi
  0 siblings, 0 replies; 2+ messages in thread
From: Al Boldi @ 2006-09-22 14:24 UTC (permalink / raw)
  To: linux-kernel

Ludovic Drolez wrote:
> Ludovic Drolez <ldrolez <at> linbox.com> writes:
> > In fact, I tested the 1st patch on our cluster (Finite elements
> > computing on 8 CPUs):
> > - Under Windows: 875 seconds
> > - Linux 2.6.16 : 1019 s
> > - Linux 2.6.16 + manual taskset : 842 s
> > - Linux 2.6.16 + Vincent's patch : 1373 s
>
> Anyone has an idea why the scheduling is poor when processes don't use all
> CPU ?
>
> In the above example, we have 4 processes on 4 processors which use about
> 40% of the CPU (computing and waiting for network packets).
> 1- If taskset is not used : CPU0 is used at 80%, and the 3 others at 30%.
> The tasks are constantly migrated between cores -> poor performance
> (1019s), Windows does better :-(
> 2- If taskset is used : All CPUs have 1 process and are used at 40%. No
> migration -> high performance (842s), better than Windows :-)
>
> I tried to play with the 'migration_cost' kernel parameter but it did not
> help. By default, on the Bi-Xeon Dual Core MB (Dell 1855),
> migration_cost=1600, and trying values up to 200000, did not improve
> performance...
>
> Any Ideas ?

Did you try PlugSched?  The spa scheds work wonders, once tuned properly.


Thanks!


--
Al


^ permalink raw reply	[flat|nested] 2+ messages in thread

* Poor scheduling when not loaded at 100% (Was: [PATCH] sched.c: Be a bit more conservative in SMP)
  2006-09-20  7:42         ` Ludovic Drolez
@ 2006-09-20 16:26           ` Ludovic Drolez
  0 siblings, 0 replies; 2+ messages in thread
From: Ludovic Drolez @ 2006-09-20 16:26 UTC (permalink / raw)
  To: linux-kernel

Ludovic Drolez <ldrolez <at> linbox.com> writes:
> In fact, I tested the 1st patch on our cluster (Finite elements computing on 8 
> CPUs):
> - Under Windows: 875 seconds
> - Linux 2.6.16 : 1019 s
> - Linux 2.6.16 + manual taskset : 842 s
> - Linux 2.6.16 + Vincent's patch : 1373 s  

Anyone has an idea why the scheduling is poor when processes don't use all CPU ?

In the above example, we have 4 processes on 4 processors which use about 40% of
the CPU (computing and waiting for network packets).
1- If taskset is not used : CPU0 is used at 80%, and the 3 others at 30%. The
tasks are constantly migrated between cores -> poor performance (1019s), Windows
does better :-(
2- If taskset is used : All CPUs have 1 process and are used at 40%. No
migration -> high performance (842s), better than Windows :-)

I tried to play with the 'migration_cost' kernel parameter but it did not help.
By default, on the Bi-Xeon Dual Core MB (Dell 1855), migration_cost=1600, and
trying values up to 200000, did not improve performance...

Any Ideas ?

  Ludovic Drolez.






^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2006-09-22 14:22 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2006-09-22 14:24 Poor scheduling when not loaded at 100% (Was: [PATCH] sched.c: Be a bit more conservative in SMP) Al Boldi
  -- strict thread matches above, loose matches on Subject: below --
2006-09-03 13:41 [PATCH] sched.c: Be a bit more conservative in SMP Vincent Pelletier
2006-09-03 17:10 ` Vincent Pelletier
2006-09-06 23:30   ` Vincent Pelletier
2006-09-19 14:06     ` Ludovic Drolez
2006-09-19 17:50       ` Antonio Vargas
2006-09-20  7:42         ` Ludovic Drolez
2006-09-20 16:26           ` Poor scheduling when not loaded at 100% (Was: [PATCH] sched.c: Be a bit more conservative in SMP) Ludovic Drolez

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox

all inboxes | Powered by JetHome®