linux-mm.kvack.org archive mirror
 help / color / mirror / Atom feed
From: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
To: YAMAMOTO Takashi <yamamoto@valinux.co.jp>
Cc: linux-mm@kvack.org, containers@lists.osdl.org,
	balbir@linux.vnet.ibm.com, xemul@openvz.org,
	nishimura@mxp.nes.nec.co.jp, menage@google.com
Subject: Re: [RFD][PATCH] memcg: Move Usage at Task Move
Date: Tue, 10 Jun 2008 17:13:48 +0900	[thread overview]
Message-ID: <20080610171348.fb7aa360.kamezawa.hiroyu@jp.fujitsu.com> (raw)
In-Reply-To: <20080610055032.A8AB25A0E@siro.lan>

On Tue, 10 Jun 2008 14:50:32 +0900 (JST)
yamamoto@valinux.co.jp (YAMAMOTO Takashi) wrote:

> >  3. Use Lazy Manner
> >       When the task moves, we can mark the pages used by it as
> >       "Wrong Charge, Should be dropped", and add them some penalty in the LRU.
> >     Pros.
> >       - no complicated ones.
> >       - the pages will be gradually moved at memory pressure.
> >     Cons.
> >       - A task's usage can exceed the limit for a while.
> >       - can't handle mlocked() memory in proper way.
> > 
> >  4. Allow Half-moved state and abandon rollback.
> >     Pros.
> >       - no complicated ones in the code.
> >     Cons.
> >       - the users will be in chaos.
> 
> how about:
> 
> 5. try to move charges as your patch does.
>    if the target cgroup's usage is going to exceed the limit,
>    try to shrink it.  if it failed, just leave it exceeded.
>    (ie. no rollback)
>    for the memory subsystem, which can use its OOM killer,
>    the failure should be rare.
> 

Hmm, allowing exceed and cause OOM kill ?

One difficult point is that the users cannot know they can move task
without any risk. How to handle the risk can be a point. 
I don't like that approarch in general because I don't like "exceed"
status. But implementation will be easy.

> > After writing this patch, for me, "3" is attractive. now.
> > (or using Lazy manner and allow moving of usage instead of freeing it.)
> > 
> > One reasone is that I think a typical usage of memory controller is
> > fork()->move->exec(). (by libcg ?) and exec() will flush the all usage.
> 
> i guess that moving long-running applications can be desirable
> esp. for not so well-designed systems.
> 

hmm, for not so well-designed systems....true.
But "5" has the same kind of risks for not so well-desgined systems ;)


Thanks,
-Kame

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

  reply	other threads:[~2008-06-10  8:13 UTC|newest]

Thread overview: 32+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2008-06-06  1:52 KAMEZAWA Hiroyuki
2008-06-10  5:50 ` YAMAMOTO Takashi
2008-06-10  8:13   ` KAMEZAWA Hiroyuki [this message]
2008-06-10 12:57     ` YAMAMOTO Takashi
2008-06-11  2:02       ` KAMEZAWA Hiroyuki
2008-06-11  3:45         ` YAMAMOTO Takashi
2008-06-11  4:08           ` KAMEZAWA Hiroyuki
2008-06-10  7:35 ` Daisuke Nishimura
2008-06-10  8:26   ` KAMEZAWA Hiroyuki
2008-06-11  3:03     ` Daisuke Nishimura
2008-06-11  3:25       ` KAMEZAWA Hiroyuki
2008-06-11  3:44         ` YAMAMOTO Takashi
2008-06-11  4:14           ` KAMEZAWA Hiroyuki
2008-06-11  4:29             ` Daisuke Nishimura
2008-06-11  4:40               ` KAMEZAWA Hiroyuki
2008-06-12  5:20             ` YAMAMOTO Takashi
2008-06-12  6:51               ` KAMEZAWA Hiroyuki
2008-06-11  7:17 ` Paul Menage
2008-06-11  7:45   ` KAMEZAWA Hiroyuki
2008-06-11  8:04     ` Paul Menage
2008-06-11  8:27       ` KAMEZAWA Hiroyuki
2008-06-11  8:48         ` Paul Menage
2008-06-12  5:08           ` KAMEZAWA Hiroyuki
2008-06-12 13:17             ` Serge E. Hallyn
2008-06-12 13:34             ` kamezawa.hiroyu
2008-06-12 21:08               ` Serge E. Hallyn
2008-06-13  0:34                 ` KAMEZAWA Hiroyuki
2008-06-13  0:41                   ` KAMEZAWA Hiroyuki
2008-06-11  8:27   ` Balbir Singh
2008-06-11 12:21     ` Daisuke Nishimura
2008-06-11 12:51     ` kamezawa.hiroyu
2008-06-11 13:13       ` Balbir Singh

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20080610171348.fb7aa360.kamezawa.hiroyu@jp.fujitsu.com \
    --to=kamezawa.hiroyu@jp.fujitsu.com \
    --cc=balbir@linux.vnet.ibm.com \
    --cc=containers@lists.osdl.org \
    --cc=linux-mm@kvack.org \
    --cc=menage@google.com \
    --cc=nishimura@mxp.nes.nec.co.jp \
    --cc=xemul@openvz.org \
    --cc=yamamoto@valinux.co.jp \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox