[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Bug#534964: Any updates on this BUG?



* Ben Hutchings <ben@decadent.org.uk> [2011-05-23 08:51:00]:

> On Mon, 2011-05-23 at 21:03 +0530, Balbir Singh wrote:
> > * Ben Hutchings <ben@decadent.org.uk> [2011-05-23 08:10:21]:
> > 
> > > On Mon, 2011-05-23 at 12:43 +0530, Balbir Singh wrote:
> > > > * Ben Hutchings <ben@decadent.org.uk> [2011-05-23 00:10:13]:
> > > > 
> > > > > On Fri, 2011-05-20 at 17:35 +0530, Balbir Singh wrote:
> > > > > > * maximilian attems <maks@debian.org> [2011-05-12 11:53:25]:
> > > > > > 
> > > > > > > On Wed, May 11, 2011 at 12:08:10PM +0000, maximilian attems wrote:
> > > > > > > > On Wed, May 11, 2011 at 11:33:31AM +0000, maximilian attems wrote:
> > > > > > > > > On Wed, May 11, 2011 at 04:43:06PM +0530, Balbir Singh wrote:
> > > > > > > > > > <balbir@linux.vnet.ibm.com> wrote:
> > > > > > > > > > > Hi, Ben,
> > > > > > > > > > >
> > > > > > > > > > > We've made significant progress in reducing the overhead of memory
> > > > > > > > > > > cgroup subsystem. I'd request you to try it and enable it. If there
> > > > > > > > > > > are some concerns, we could always address them. There are more
> > > > > > > > > > > interesting changes on the way as well.
> > > > > > > > 
> > > > > > > > could you point to the relevant commits?
> > > > > > > 
> > > > > > > so unless this information comes along soon,
> > > > > > > I'll revert my change.
> > > > > > >
> > > > > > 
> > > > > > I sent some information earlier, I hope it was useful. BTW, I don't
> > > > > > see memory cgroups enabled in debian experimental (changelog), should I be looking
> > > > > > elsewhere? 
> > > > > 
> > > > > They're enabled in 2.6.39-1 (in sid).  However, I actually meant them to
> > > > > be disabled by default, and that will be done in 2.6.39-2.
> > > > >
> > > > 
> > > > Oops, but why do you want to disable them by default? Other distros
> > > > have them enabled, disabled implies more boot options required, lesser
> > > > testing of the feature and feedback. Is there a strong reason to
> > > > change it? 
> > > 
> > > Yes, because it currently (2.6.39) requires extra page tables that take
> > > time and memory.  I understand this is supposed to change soon, and we
> > > can reconsider then.
> > >
> > 
> > Sorry, I don't understand the extra page tables remark, that changed
> > way back in 2.6.30/1 timeframe.
> 
> There's an array of page_cgroup structures, mirroring the page
> structures.
> 
> > Could you please elaborate, so that I
> > can clarify the correct concern. 
> 
> http://lwn.net/SubscriberLink/443241/75143317a94d9752/
>

Yes, but that is not page tables, that is struct page mirroring. That
uses less than 1% (about 0.7%) of the total memory, is that too bad?



-- 
	Three Cheers,
	Balbir



Reply to: