Re: [PATCH] cfq-iosched: reduce write depth only if sync was delayed

[Jens Axboe <jens.axboe@oracle.com>]

On Wed, Dec 09 2009, Jeff Moyer wrote:
> Corrado Zoccolo <czoccolo@gmail.com> writes:
> 
> > Hi Jeff,
> > On Tue, Dec 8, 2009 at 7:00 PM, Jeff Moyer <jmoyer@redhat.com> wrote:
> >> Corrado Zoccolo <czoccolo@gmail.com> writes:
> >>
> >>> The introduction of ramp-up formula for async queue depths has
> >>> slowed down dirty page reclaim, by reducing async write performance.
> >>> This patch makes sure the formula kicks in only when sync request
> >>> was recently delayed.
> >>> @@ -3706,7 +3707,7 @@ static void *cfq_init_queue(struct request_queue *q)
> >>>       cfqd->cfq_latency = 1;
> >>>       cfqd->cfq_group_isolation = 0;
> >>>       cfqd->hw_tag = -1;
> >>> -     cfqd->last_end_sync_rq = jiffies;
> >>> +     cfqd->last_delayed_sync = jiffies - HZ;
> >>
> >> So, umm, what's that about?
> >
> > Previously, when cfq started, it started in a state where writes were
> > completely throttled. Now, we optimistically prefer to start with a
> > reasonable max depth (10)
> 
> OK.  Can we put a comment in there and change the initialization to
> cfq_slice_sync * 10?

Agree, that would be MUCH easier to understand.

-- 
Jens Axboe