[lkml]   [2015]   [Nov]   [18]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
SubjectRe: [PATCH 13/14] mm: memcontrol: account socket memory in unified hierarchy memory controller
On Wed, Nov 18, 2015 at 05:22:56PM +0100, Michal Hocko wrote:
> On Mon 16-11-15 13:18:10, Johannes Weiner wrote:
> > What load would you test and what would be the baseline to compare it
> > to?
> It seems like netperf with a stream load running in a memcg with no
> limits vs. in root memcg (and no other cgroups) should give at least a
> hint about the runtime overhead, no?

Comparing root vs. dedicated group generally doesn't make sense since
you either need containment or you don't. It makes more sense to test
both times inside a memory-controlled cgroup, one with a regular boot,
one with cgroup.memory=nosocket.

So I ran perf record -g -a netperf -t TCP_STREAM multiple times inside
a memory-controlled cgroup, but mostly mem_cgroup_charge_skmem() does
not show up in the profile at all. Once it was there with 0.00%.

I ran another test that downloads the latest kernel image from at 13MB/s (on my i5 laptop) and it looks like this:

0.02% 0.01% irq/44-iwlwifi [kernel.kallsyms] [k] mem_cgroup_charge_skmem

The runs vary too much for this to be measurable in elapsed time.

 \ /
  Last update: 2015-11-18 23:21    [W:0.107 / U:3.100 seconds]
©2003-2018 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site