Re: [RFC V2 PATCH 1/5] cfq-iosched: adapt slice to number of processes doing I/O

[Corrado Zoccolo <czoccolo@gmail.com>]

Hi Jeff,
On Wed, Oct 21, 2009 at 6:36 PM, Jeff Moyer <jmoyer@redhat.com> wrote:
> Corrado Zoccolo <czoccolo@gmail.com> writes:
>
> OK.  The idea was to try to make the thing a bit more digestable.
> Honestly, I had to re-write it to figure out what it was doing.  Could
> you rework it so the logic is more obvious to others and still correct?

What about:
/* interested queues (we consider only the ones with the same priority class) */
unsigned iq = cfq_get_avg_queues(cfqd, cfq_class_rt(cfqq));
unsigned sync_slice = cfqd->cfq_slice[1];
unsigned expected_latency = sync_slice * iq;
if (expected_latency > target_latency) {
	unsigned base_low_slice = 2 * cfqd->cfq_slice_idle;
	/* scale low_slice according to IO priority and sync vs async */
	unsigned low_slice = min(slice,  base_low_slice * slice / sync_slice);
	/* the adapted slice value is scaled to fit all iqs into the target latency */
	slice = max(slice * cfq_target_latency / expected_latency, low_slice);
}

BTW, the other complex part of the patch, i.e. the definition of
cfq_get_avg_queues, could be simplified if we hard-code the divisor,
that is currently defined outside to allow for some experimentation.

Thanks for your comments
Corrado

>
> Thanks!
> Jeff
>



-- 
__________________________________________________________________________

dott. Corrado Zoccolo                          mailto:czoccolo@gmail.com
PhD - Department of Computer Science - University of Pisa, Italy
--------------------------------------------------------------------------