lkml.org 
[lkml]   [2012]   [Sep]   [13]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
SubjectRe: [RFC] sched: nohz_idle_balance
From
Date
On Thu, 2012-09-13 at 10:45 +0200, Peter Zijlstra wrote:
> On Thu, 2012-09-13 at 10:37 +0200, Vincent Guittot wrote:
> > > I think you need to present numbers showing benefit. Crawling all over
> > > a mostly idle (4096p?) box is decidedly bad thing to do.
>
> Yeah, but we're already doing that anyway.. we know nohz idle balance
> doesn't scale. Venki and Suresh were working on that, but with Venki
> switching jobs this work got dropped.
>
> I did talk to Suresh about it a week or so ago.. I think he was going to
> look at it again.

As a reminder to Suresh.. please also consider calc_load_{idle,idx} in
this work, its another nohz 'feature' that doesn't scale for pretty much
the same reason.

That is, I'm fine with the initial patches not actually fixing that, but
the structure put in place for the ILB should allow for it.


\
 
 \ /
  Last update: 2012-09-13 11:41    [W:0.038 / U:0.276 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site