Re: [RFI] Shared accounting for memory resource controller

[Balbir Singh]

* KAMEZAWA Hiroyuki <kamezawa.hiroyu@xxxxxxxxxxxxxx> [2009-04-08 16:18:24]:

> On Wed, 8 Apr 2009 12:41:15 +0530
> Balbir Singh <balbir@xxxxxxxxxxxxxxxxxx> wrote:
> 
> > * KAMEZAWA Hiroyuki <kamezawa.hiroyu@xxxxxxxxxxxxxx> [2009-04-08 16:07:33]:
> > 1. First our rss in memory.stat is confusing, we should call it anon
> > RSS
> ok. but ....changing current interface ?
> 

No, lets just a new field called file_rss and make sure the
documentation reflects the correct information.

> > 2. We need to add file rss, this is sort of inline with the
> > information we export per process file_rss and anon_rss
> 
> maybe good. *but* active/incative ratio in lru file cache is good estimation for this.

Active/Inactive tell us about frequently a page is referenced rather
than what is mapped and what is not. We could get very bad results if
we make an assumption.

> 
> > 3. Using the above, we can then try to (using an algorithm you
> > proposed), try to do some work for figuring out the shared percentage.
> > 
> This is the point. At last. Why "# of shared pages" is important ?
> 

I posted this in my motivation yesterday. # of shared pages can help
plan the system better and the size of the cgroup. A cgroup might have
small usage_in_bytes but large number of shared pages. We need a
metric that can help figure out the fair usage of the cgroup.

> I wonder it's better to add new stat file as memory.cacheinfo which helps
> following kind of commands.
> 
>   #cacheinfo /cgroups/memory/group01/
>        /usr/lib/libc.so.1     30pages
>        /var/log/messages      1 pages
>        /tmp/xxxxxx            20 pages
>        .....
>        .....
>

But, what I need at the moment is shared usage information.

-- 
	Balbir
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/