[lkml]   [2011]   [Nov]   [24]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
SubjectRe: [PATCH v2 14/14] Change CPUACCT to default n
On 11/24/2011 11:24 AM, Peter Zijlstra wrote:
> On Mon, 2011-11-21 at 10:59 +0900, KAMEZAWA Hiroyuki wrote:
>> Could you share your analysis why cpuacct cgroup is slow ?
> It adds accounting to all scheduler hot paths, accounting that is mostly
> duplicate of accounting already done. It also does another cgroup
> hierarchy walk, separate from existing controllers, hitting all cold
> cachelines again.
> IOW, it sucks chunks and I should never have allowed it to be merged.

OTOH, if the use case for it includes separating processes for the cpu
and cpuacct cgroups in an independent manner - which apparently it does,
I've just learned, there isn't much we can do except try to make it cheaper.

The direction I am going right now wrt this is to clean this up and
reduce the impact of it when the root cgroup is the only one active. All
other users are using the functionality, so let them pay...

 \ /
  Last update: 2011-11-24 17:11    [W:0.052 / U:2.780 seconds]
©2003-2018 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site