* [Cluster-devel] [GFS2 PATCH 0/5] GFS2: Bitmap search optimizations
@ 2013-09-11 18:43 Bob Peterson
2013-09-12 8:30 ` Steven Whitehouse
0 siblings, 1 reply; 2+ messages in thread
From: Bob Peterson @ 2013-09-11 18:43 UTC (permalink / raw)
To: cluster-devel.redhat.com
This set of 5 patches increases the speed of GFS2's bitmap search
algorithms. To test the performance, I added instrumentation to
function gfs2_inplace_reserve to report the amount of time spent
there. With the "stock" gfs2, I got 64586490 microseconds while
running postmark with 100000 transactions and 100000 files.
With the patched gfs2, I got 49441435, which is nearly 24% faster.
Without the 5 GFS2 patches, postmark gave this result:
852.13 megabytes written (970.61 kilobytes per second)
With the 5 patches GFS2, I got:
852.13 megabytes written (979.32 kilobytes per second)
So overall, postmark performance was only 1% faster.
Using straight dd copying 1MB files, I got 10% faster. Obviously
mileage will vary: my hardware isn't the fastest.
The patches are as follows:
[PATCH 1/5] GFS2: optimize rbm_from_block wrt bi_start
[PATCH 2/5] GFS2: introduce bi_blocks for optimization
[PATCH 3/5] GFS2: Introduce rbm field bii
[PATCH 4/5] GFS2: new function gfs2_rbm_incr
[PATCH 5/5] GFS2: Do not reset flags on active reservations
Regards,
Bob Peterson
Red Hat File Systems
^ permalink raw reply [flat|nested] 2+ messages in thread
* [Cluster-devel] [GFS2 PATCH 0/5] GFS2: Bitmap search optimizations
2013-09-11 18:43 [Cluster-devel] [GFS2 PATCH 0/5] GFS2: Bitmap search optimizations Bob Peterson
@ 2013-09-12 8:30 ` Steven Whitehouse
0 siblings, 0 replies; 2+ messages in thread
From: Steven Whitehouse @ 2013-09-12 8:30 UTC (permalink / raw)
To: cluster-devel.redhat.com
Hi,
This looks very good. I'll queue it just as soon as the merge window is
over so that we don't have issues around -next,
Steve.
On Wed, 2013-09-11 at 13:43 -0500, Bob Peterson wrote:
> This set of 5 patches increases the speed of GFS2's bitmap search
> algorithms. To test the performance, I added instrumentation to
> function gfs2_inplace_reserve to report the amount of time spent
> there. With the "stock" gfs2, I got 64586490 microseconds while
> running postmark with 100000 transactions and 100000 files.
> With the patched gfs2, I got 49441435, which is nearly 24% faster.
> Without the 5 GFS2 patches, postmark gave this result:
> 852.13 megabytes written (970.61 kilobytes per second)
> With the 5 patches GFS2, I got:
> 852.13 megabytes written (979.32 kilobytes per second)
> So overall, postmark performance was only 1% faster.
> Using straight dd copying 1MB files, I got 10% faster. Obviously
> mileage will vary: my hardware isn't the fastest.
>
> The patches are as follows:
>
> [PATCH 1/5] GFS2: optimize rbm_from_block wrt bi_start
> [PATCH 2/5] GFS2: introduce bi_blocks for optimization
> [PATCH 3/5] GFS2: Introduce rbm field bii
> [PATCH 4/5] GFS2: new function gfs2_rbm_incr
> [PATCH 5/5] GFS2: Do not reset flags on active reservations
>
> Regards,
>
> Bob Peterson
> Red Hat File Systems
>
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2013-09-12 8:30 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2013-09-11 18:43 [Cluster-devel] [GFS2 PATCH 0/5] GFS2: Bitmap search optimizations Bob Peterson
2013-09-12 8:30 ` Steven Whitehouse
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).