[Ksummit-2012-discuss] [ATTEND] writeback and kernel testing

Fengguang Wu fengguang.wu at intel.com
Tue Jun 19 13:10:16 UTC 2012


On Tue, Jun 19, 2012 at 02:42:05PM +0200, Johannes Weiner wrote:
> On Mon, Jun 18, 2012 at 06:18:49PM +0800, Fengguang Wu wrote:
> > On Mon, Jun 18, 2012 at 02:57:52PM +0530, Balbir Singh wrote:
> > > On Sun, Jun 17, 2012 at 8:47 AM, Fengguang Wu <fengguang.wu at intel.com> wrote:
> > > > On Sun, Jun 17, 2012 at 08:03:15AM +0530, Balbir Singh wrote:
> > > >> On Sat, Jun 16, 2012 at 6:14 PM, Fengguang Wu <fengguang.wu at intel.com> wrote:
> > > >> > Greetings,
> > > >> >
> > > >> > I'd like to attend this year's kernel summit.
> > > >> >
> > > >> > I may talk about the technical challenges and trade-offs on writeback and memcg
> > > >> > and IO controllers with anyone interested, perhaps in some breakout session.
> > > >> >
> > > >>
> > > >> Do you have a reference for the tests you've mentioned. Do you have a
> > > >> brief summary of the challenges and trade-offs?
> > > >
> > > > - memcg with dirty limits
> > > >
> > > >  The basic agreements are, we'll do per-memcg dirty limit, but set the
> > > >  limits reasonably high and do not expose interfaces to user space.
> > > >
> > > 
> > > No interfaces in user-space, how is it expected to work via the global limits?
> > 
> > Yeah, we could implicitly tie the per-memcg dirty limits to the global one.
> > For example, do 40% memcg dirty ratio when the global ratio is 20%.
> > 
> > At the same time, we'll need to improve the page reclaim code so that
> > it works well even when a LRU list has 40% dirty pages. I believe this
> > is an achievable goal when we put dirty pages to a standalone LRU list.
> 
> It was dropped because nobody considered it urgent/important enough to
> sacrifice a pageflag for it.  But I still think this is a good idea.

Fair enough. Actually I was just proposing that and treat it a
mid-term goal :)

> The problem is worse with my refault distance-based active list
> sizing, because it allows the inactive list to shrink to a minimum,
> which may be all dirty (with the zone's clean pages sitting on the
> active list), so I wanted to repropose this change in the near future
> anyway.

Okay!

> Minchan also just suggested to have an easy-to-reclaim LRU list, which
> not just splits out dirty pages, but those marked for reclaim by
> fadvise/madvise/reclaim-during-writeback etc.

Oh, my expected LRU_DIRTY would also contain the writeback pages :)

Thanks,
Fengguang


More information about the Ksummit-2012-discuss mailing list