Messages in this thread | | | Date | Sun, 17 Oct 2010 17:39:33 +0530 | From | Balbir Singh <> | Subject | Re: [RFC tg_shares_up improvements - v1 00/12] [RFC tg_shares_up - v1 00/12] Reducing cost of tg->shares distribution |
| |
* Peter Zijlstra <peterz@infradead.org> [2010-10-17 11:38:51]:
> On Sun, 2010-10-17 at 10:54 +0530, Balbir Singh wrote: > > > > The results do sound interesting, could I recommend that you try to > > get 1000 cgroups or so, with half of them as active and half inactive. > > I'd also suggest tracking system and user time at the cgroup level. > > Sure, but do keep in mind that 500 active cgroups will always utterly > suck no matter what you do, it means we do indeed need to iterate 500 > cgroups for a lot of scheduler operations. >
I agree in general, but in my case I've seen nasty things even when the groups were empty.
-- Three Cheers, Balbir
| |