Re: RFC on io-stalls patch

[Jens Axboe <axboe@suse.de>]

On Tue, Jul 15 2003, Andrea Arcangeli wrote:
> On Tue, Jul 15, 2003 at 11:18:38AM +0200, Jens Axboe wrote:
> > On Tue, Jul 15 2003, Jens Axboe wrote:
> > > > BTW, the contest run times vary pretty wildy.  My 3 compiles with
> > > > io_load running on 2.4.21 were 603s, 443s and 515s.  This doesn't make
> > > > the average of the 3 numbers invalid, but we need a more stable metric.
> > > 
> > > Mine are pretty consistent [1], I'd suspect that it isn't contest but your
> > > drive tcq skewing things. But it would be nice to test with other things
> > > as well, I just used contest because it was at hand.
> > 
> > Oh and in the same spirit, I'll do the complete runs on an IDE drive as
> > well. Sometimes IDE vs SCSI shows the funniest things.
> 
> this is the first suspect IMHO too. Especially given the way that SCSI
> releases the requests. 

2.4.21 + 2.4.22-pre5 + 2.4.22-pre5-axboe is running/pending on IDE now
here.

> One more thing: unlike my patch where I forced all drivers to support
> elevator-lowlatency (either that or not compile at all), Chris made it
> optional when he pushed it into mainline, so now the device driver has
> to call blk_queue_throttle_sectors(q, 1) to enable it. Otherwise it'll
> run like 2.4.21.

Right, but both IDE and SCSI sets it. IDE is still the best to bench io
scheduler performance on, though.

-- 
Jens Axboe