Re: [RFC PATCHSET v2] sched/fair: fix load balancer behavior when cgroup is in use

From: Tejun Heo
Date: Fri May 05 2017 - 09:28:57 EST


Hello, Vincent.

On Fri, May 05, 2017 at 10:46:53AM +0200, Vincent Guittot wrote:
> schbench results looks better with this version
> Latency percentiles (usec)
> 50.0000th: 212
> 75.0000th: 292
> 90.0000th: 385
> 95.0000th: 439
> *99.0000th: 671
> 99.5000th: 7992
> 99.9000th: 12176
> min=0, max=14855
>
> p99 is back to a normal value but p99.5 stays higher than mainline

By how much and is that with the weight adjustment on the cgroup? I
can't get reliable numbers on p99.5 and beyond on my test setup. I
ordered the hikey board and will try to replicate your setup once it
arrives.

> I have also checked load_avg and runnable_load_avg value and there is
> something incorrect. I will provide details on the related patch

Sure, will respond there.

Thanks.

--
tejun