[RFC] writeback and cgroup

Tejun Heo tj at kernel.org
Mon Apr 23 16:04:54 UTC 2012


Hello, Vivek.

On Mon, Apr 23, 2012 at 08:30:11AM -0400, Vivek Goyal wrote:
> On Fri, Apr 20, 2012 at 02:33:01PM -0700, Tejun Heo wrote:
> > On Fri, Apr 20, 2012 at 03:29:30PM -0400, Vivek Goyal wrote:
> > > I am personally is not too excited about the case of putting async IO
> > > in separate groups due to the reason that async IO of one group will
> > > start impacting latencies of sync IO of another group and in practice
> > > it might not be desirable. But there are others who have use cases for
> > > separate async IO queue. So as long as switch is there to change the
> > > behavior, I am not too worried.
> > 
> > Why not just fix cfq so that it prefers groups w/ sync IOs?
> 
> Yes that could possibly be done but now that's change of requirements. Now
> we are saying that I want one buffered write to go faster than other
> buffered write only if there is no sync IO present in any of the groups.

It's a scheduling decision and the resource split may or may not be
about latency (the faster part).  We're currently just shoving all
asyncs into the root group and preferring sync IOs in general.  The
other end would be keeping them completely siloed and not caring about
[a]sync across different cgroups.  My point is that managing async IOs
per cgroup doesn't mean we can't prioritize sync IOs in general.

Thanks.

-- 
tejun


More information about the Containers mailing list