Re: [Lse-tech] Re: [PATCH][RFC] Proposal For A More Scalable Scheduler ...

[Mike Kravetz]

On Tue, Oct 30, 2001 at 05:06:16PM -0800, Davide Libenzi wrote:
> On Tue, 30 Oct 2001, Mike Kravetz wrote:
> > --------------------------------------------------------------------
> > Chat - VolanoMark simulator.  Result is a measure of throughput.
> >        Higher is better.
> > --------------------------------------------------------------------
> 
> How does this test work exactly ?
> Did You measure the run queue length while this was running ?
> I'm going to modify LatSched to output other info like rqlen and
> tsk->counter at switch time.
> I'd like to have a  schedcnt  dump while this test is running.
> Anyway this test shows that what i'm doing now ( working on the balancing
> schemes ) is not wasted time :)

Take a look at our OLS presentation slides starting at:

http://lse.sourceforge.net/scheduling/ols2001/img50.htm

This is basically a message passing (via sockets) benchmark with
high thread counts/runqueue lengths.

I'll kick off the all important 'kernel build benchmark' and have
some results tomorrow.

-- 
Mike