On Thu, 19 May 2011 20:24:51 -0700
Ying Han <
yinghan@google.com> wrote:
> The new API exports numa_maps per-memcg basis. This is a piece of useful
> information where it exports per-memcg page distribution across real numa
> nodes.
>
> One of the usecase is evaluating application performance by combining this
> information w/ the cpu allocation to the application.
>
> The output of the memory.numastat tries to follow w/ simiar format of numa_maps
> like:
>
> total=<total pages> N0=<node 0 pages> N1=<node 1 pages> ...
> file=<total file pages> N0=<node 0 pages> N1=<node 1 pages> ...
> anon=<total anon pages> N0=<node 0 pages> N1=<node 1 pages> ...
> unevictable=<total anon pages> N0=<node 0 pages> N1=<node 1 pages> ...
>
> And we have per-node:
> total = file + anon + unevictable
>
> $ cat /dev/cgroup/memory/memory.numa_stat
> total=250020 N0=87620 N1=52367 N2=45298 N3=64735
> file=225232 N0=83402 N1=46160 N2=40522 N3=55148
> anon=21053 N0=3424 N1=6207 N2=4776 N3=6646
> unevictable=3735 N0=794 N1=0 N2=0 N3=2941
>
> change v4..v3:
> 1. add per-node "unevictable" value.
> 2. change the functions to be static.
>
> change v3..v2:
> 1. calculate the "total" based on the per-memcg lru size instead of rss+cache.
> this makes the "total" value to be consistant w/ the per-node values follows
> after.
>
> change v2..v1:
> 1. add also the file and anon pages on per-node distribution.
>
> Signed-off-by: Ying Han <
yinghan@google.com>