Re: [BENCHMARK] scheduler tunables with contest - max_timeslice

Robert Love (rml@tech9.net)
18 Dec 2002 20:40:06 -0500


On Wed, 2002-12-18 at 20:31, Con Kolivas wrote:

> Using the osdl (http://www.osdl.org) resources provided to me I'm running a
> series of contest benchmarks on 2.5.52-mm1 and modifying the scheduler tunables
> as provided by RML's patch. SMP used to minimise how long it will take me to do
> these. This is the first in the series and I've run a range of max_timeslices
> (default is 300ms; range 100-400):

Thanks, Con.

> I will continue to do these with some of the other scheduler tunables. I will
> need recommendations if anyone is interested in further resolution testing than
> that I'm currently doing.

Some ideas...

Try child_penalty=50,75 (default is 95%)

Try max_sleep_avg=500,1000,4000 (default is 2000ms)

Try prio_bonus_ratio=0,10,30,50 (default is 25%)

Try starvation_limit=1000 (default is 2000ms)

Some of these are just academic, although I am curious about
child_penalty and starvation_limit. In other cases, we can assuredly
improve interactivity but we have to worry about throughput and
starvation. Nonetheless, I am curious as to what you will find. :)

Robert Love

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/