Re: dramatic I/O slowdown after upgrading 2.6.38->3.0+
From: Jan Kara
Date: Tue Apr 10 2012 - 15:51:48 EST
On Wed 11-04-12 00:55:44, Suresh Jayaraman wrote:
> On 04/10/2012 08:43 PM, Jan Kara wrote:
> > On Tue 10-04-12 10:00:38, Michael Tokarev wrote:
> >> On 10.04.2012 06:26, Dave Chinner wrote:
> >>
> >>> Barriers. Turn them off, and see if that fixes your problem.
> >>
> >> Thank you Dave for a hint. And nope, that's not it, not at all... ;)
> >> While turning off barriers helps a tiny bit, to gain a few %% from
> >> the huge slowdown, it does not cure the issue.
> >>
> >> Meanwhile, I observed the following:
> >>
> >> 1) the issue persists on more recent kernels too, I tried 3.3
> >> and it is also as slow as 3.0.
> >>
> >> 2) at least 2.6.38 kernel works fine, as fast as 2.6.32, I'll
> >> try 2.6.39 next.
> >>
> >> I updated $subject accordingly.
> >>
> >> 3) the most important thing I think: this is general I/O speed
> >> issue. Here's why:
> >>
> >> 2.6.38:
> >> # dd if=/dev/sdb of=/dev/null bs=1M iflag=direct count=100
> >> 100+0 records in
> >> 100+0 records out
> >> 104857600 bytes (105 MB) copied, 1.73126 s, 60.6 MB/s
> >>
> >> 3.0:
> >> # dd if=/dev/sdb of=/dev/null bs=1M iflag=direct count=100
> >> 100+0 records in
> >> 100+0 records out
> >> 104857600 bytes (105 MB) copied, 29.4508 s, 3.6 MB/s
> >>
> >> That's about 20 times difference on direct read from the
> >> same - idle - device!!
> > Huh, that's a huge difference for such a trivial load. So we can rule out
> > filesystems, writeback, mm. I also wouldn't think it's IO scheduler but
> > you can always check by comparing dd numbers after
> > echo none >/sys/block/sdb/queue/scheduler
>
> s/none/noop
> you meant noop, of course?
Yeah. Thanks for correction!
Honza
--
Jan Kara <jack@xxxxxxx>
SUSE Labs, CR
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/