Prev: oom: fix the unsafe usage of badness() in proc_oom_score()
Next: drivers/char/amiserial.c: Add missing local_irq_restore
From: Jeff Moyer on 5 Apr 2010 14:40
tytso(a)mit.edu writes:

> On Thu, Apr 01, 2010 at 03:04:54PM -0400, Jeff Moyer wrote:
>>
>> So, I'm submitting this patch for comments and testing. I have a
>> similar patch for jbd that I will submit if folks agree that this is a
>> good idea.
>
> Added to the ext4 patch queue.
>
> What benchmark were you using to test small file writes? This looks
> good to me as well, but we might want to do some extra benchmarking
> just to be sure we're not accidentally introducing a performance
> regression.

iozone showed regressions for write and re-write in runs that include
fsync timings for small files (<8MB). Here's the command line used for
testing:

iozone -az -n 4k -g 2048m -y 1k -q 1m -e

I also ran fs_mark using the following command line:

fs_mark -S 1 -D 100 -N 1000 -d /mnt/test/fs_mark -s 65536 -t 1 -w 4096

I'll let you know if there are any regressions caused by this patch in
any of our other testing.

Cheers,
Jeff
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
From: tytso on 5 Apr 2010 15:50
On Mon, Apr 05, 2010 at 02:36:07PM -0400, Jeff Moyer wrote:
> >
> > What benchmark were you using to test small file writes? This looks
> > good to me as well, but we might want to do some extra benchmarking
> > just to be sure we're not accidentally introducing a performance
> > regression.
>
> iozone showed regressions for write and re-write in runs that include
> fsync timings for small files (<8MB). Here's the command line used for
> testing:
>
> iozone -az -n 4k -g 2048m -y 1k -q 1m -e

iozone is showing performance regressions or performance improvements?
I thought the point of this patch was to improve iozone benchmarks?

- Ted
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
From: Jeff Moyer on 5 Apr 2010 16:40
tytso(a)mit.edu writes:

> On Mon, Apr 05, 2010 at 02:36:07PM -0400, Jeff Moyer wrote:
>> >
>> > What benchmark were you using to test small file writes? This looks
>> > good to me as well, but we might want to do some extra benchmarking
>> > just to be sure we're not accidentally introducing a performance
>> > regression.
>>
>> iozone showed regressions for write and re-write in runs that include
>> fsync timings for small files (<8MB). Here's the command line used for
>> testing:
>>
>> iozone -az -n 4k -g 2048m -y 1k -q 1m -e
>
> iozone is showing performance regressions or performance improvements?
> I thought the point of this patch was to improve iozone benchmarks?

Sorry, Ted, what I meant to say was that iozone showed differences
between deadline and cfq, where cfq's performance was much worse than
deadline's.

Thanks!
Jeff
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
From: Jeff Moyer on 5 Apr 2010 16:50
Jeff Moyer <jmoyer(a)redhat.com> writes:

> tytso(a)mit.edu writes:
>
>> On Mon, Apr 05, 2010 at 02:36:07PM -0400, Jeff Moyer wrote:
>>> >
>>> > What benchmark were you using to test small file writes? This looks
>>> > good to me as well, but we might want to do some extra benchmarking
>>> > just to be sure we're not accidentally introducing a performance
>>> > regression.
>>>
>>> iozone showed regressions for write and re-write in runs that include
>>> fsync timings for small files (<8MB). Here's the command line used for
>>> testing:
>>>
>>> iozone -az -n 4k -g 2048m -y 1k -q 1m -e
>>
>> iozone is showing performance regressions or performance improvements?
>> I thought the point of this patch was to improve iozone benchmarks?
>
> Sorry, Ted, what I meant to say was that iozone showed differences
> between deadline and cfq, where cfq's performance was much worse than
> deadline's.

And to be 100% clear, with the patch, the performance differences
between deadline and cfq were in the noise.

Cheers,
Jeff
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
From: tytso on 5 Apr 2010 17:10
On Mon, Apr 05, 2010 at 04:41:48PM -0400, Jeff Moyer wrote:
> > Sorry, Ted, what I meant to say was that iozone showed differences
> > between deadline and cfq, where cfq's performance was much worse than
> > deadline's.
>
> And to be 100% clear, with the patch, the performance differences
> between deadline and cfq were in the noise.

Right, and since most people don't actually change the I/O scheduler
from cfq, this is basically a performance improvement patch, which is
how I'm going to describe it. :-)

- Ted
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
First  |  Prev  |  Next  |  Last
Pages: 1 2 3
Prev: oom: fix the unsafe usage of badness() in proc_oom_score()
Next: drivers/char/amiserial.c: Add missing local_irq_restore