From mboxrd@z Thu Jan 1 00:00:00 1970 From: Andrea Righi Subject: Re: [RFC PATCH 0/5] memcg: VM overcommit accounting and handling Date: Tue, 10 Jun 2008 10:30:29 +0200 (MEST) Message-ID: <484E3BA4.3000606@gmail.com> References: <1213054383-18137-1-git-send-email-righi.andrea@gmail.com> <20080610091439.04061da9.kamezawa.hiroyu@jp.fujitsu.com> <484E0D73.6010609@linux.vnet.ibm.com> <484E32B9.4020606@openvz.org> Reply-To: righi.andrea@gmail.com Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <484E32B9.4020606@openvz.org> Sender: linux-kernel-owner@vger.kernel.org To: Pavel Emelyanov , balbir@linux.vnet.ibm.com, KAMEZAWA Hiroyuki Cc: menage@google.com, kosaki.motohiro@jp.fujitsu.com, linux-kernel@vger.kernel.org, containers@lists.osdl.org List-Id: containers.vger.kernel.org Pavel Emelyanov wrote: > Balbir Singh wrote: >> KAMEZAWA Hiroyuki wrote: >>> On Tue, 10 Jun 2008 01:32:58 +0200 >>> Andrea Righi wrote: >>> >>>> Provide distinct cgroup VM overcommit accounting and handling using the memory >>>> resource controller. >>>> >>> Could you explain the benefits of this even when we have memrlimit controller ? >>> (If unsure, see 2.6.26-rc5-mm1 and search memrlimit controller.) >>> >>> And this kind of virtual-address-handling things should be implemented on >>> memrlimit controller (means not on memory-resource-controller.). >>> It seems this patch doesn't need to handle page_group. >>> >>> Considering hierarchy, putting several kinds of features on one controller is >>> not good, I think. Balbir, how do you think ? >>> >> I would tend to agree. With the memrlimit controller, can't we do this in user >> space now? Figure out the overcommit value and based on that setup the memrlimit? > > I also agree with Balbir and Kamezawa. Separate controller for VM (i.e. vma-s > lengths) is more preferable, rather than yet another fancy feature on top of > the existing rss one. > Yep! it seems I totally miss the memrlimit controller. I was trying to implement pretty the same functionalities, using a different approach. However, I agree that a separate controller seems to be a better solution. Thank you all for pointing in the right direction. I'll test memrlimit controller and give a feedback. -Andrea