All of lore.kernel.org
 help / color / mirror / Atom feed
* Re: observed significant performance improvement using "delaylog" in
@ 2010-08-12 10:46 Khelben Blackstaff
  2010-08-12 19:05 ` Michael Monnerie
  0 siblings, 1 reply; 11+ messages in thread
From: Khelben Blackstaff @ 2010-08-12 10:46 UTC (permalink / raw)
  To: xfs

> I'm pleased to state that using the experimental "delaylog" mount option 
> (in vanilla linux-2.6.35) we measured a 17% performance increase
> for our benchmark scenario. (Other mount-options in use both before
> and after the "delaylog" option: noatime,nodiratime,nobarrier)

> That's a lot given that XFS was the fastest performing file-system
> for this application already.

> So thanks to all contributing developers for this significant optimization!

I did a benchmark (a simple copy/untar not something fancy) some time ago
and also noticed great improvement.

Here is my post with the results of the benchmark.
http://lordkhelben.wordpress.com/2010/07/08/xfs-delayed-logging/

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

^ permalink raw reply	[flat|nested] 11+ messages in thread

* Re: observed significant performance improvement using "delaylog" in
  2010-08-12 10:46 observed significant performance improvement using "delaylog" in Khelben Blackstaff
@ 2010-08-12 19:05 ` Michael Monnerie
  2010-08-12 21:46   ` Peter Niemayer
  0 siblings, 1 reply; 11+ messages in thread
From: Michael Monnerie @ 2010-08-12 19:05 UTC (permalink / raw)
  To: xfs


[-- Attachment #1.1: Type: Text/Plain, Size: 679 bytes --]

On Donnerstag, 12. August 2010 Khelben Blackstaff wrote:
> Here is my post with the results of the benchmark.
> http://lordkhelben.wordpress.com/2010/07/08/xfs-delayed-logging/
 
Wow, BTRFS rocks. I guess I'll use it once it's released. But I'm 
stunned that XFS is that much slower than ext4 in many tests.

-- 
mit freundlichen Grüssen,
Michael Monnerie, Ing. BSc

it-management Internet Services
http://proteger.at [gesprochen: Prot-e-schee]
Tel: 0660 / 415 65 31

****** Aktuelles Radiointerview! ******
http://www.it-podcast.at/aktuelle-sendung.html

// Wir haben im Moment zwei Häuser zu verkaufen:
// http://zmi.at/langegg/
// http://zmi.at/haus2009/

[-- Attachment #1.2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 198 bytes --]

[-- Attachment #2: Type: text/plain, Size: 121 bytes --]

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

^ permalink raw reply	[flat|nested] 11+ messages in thread

* Re: observed significant performance improvement using "delaylog" in
  2010-08-12 19:05 ` Michael Monnerie
@ 2010-08-12 21:46   ` Peter Niemayer
  2010-08-13  9:56     ` Stan Hoeppner
  0 siblings, 1 reply; 11+ messages in thread
From: Peter Niemayer @ 2010-08-12 21:46 UTC (permalink / raw)
  To: linux-xfs

On 08/12/2010 09:05 PM, Michael Monnerie wrote:
> On Donnerstag, 12. August 2010 Khelben Blackstaff wrote:
>> Here is my post with the results of the benchmark.
>> http://lordkhelben.wordpress.com/2010/07/08/xfs-delayed-logging/
>
> Wow, BTRFS rocks.

Be sure to measure your specific use-case before jumping
to conclusions.

With our application, for example, Btrfs performed exceptionally
bad - about 4 times(!) as slow as XFS.

Then again, there are some use-cases where even older
file-systems like reiser3 excel (e.g. storing files for
cyrus imapd).


> But I'm stunned that XFS is that much slower than ext4 in many tests.

Again, it all depends on the use-case. For us, ext4
performs good (when used with all kinds of performance-enhancing,
safety-reducing mount-options), but not as good as XFS.

To me, as of today, XFS' big strength is performing good to
excellent (while not always better than all other file-systems)
in many use-cases - without worries about instability or immaturity.


One thing, I guess, is for sure: Every file-system will require
continued development to stay competitive.

SSDs, for example, are just beginning to get used appropriately
by modern file-systems. There's plenty of opportunity left to
optimize for them.

And once that is done, there may be yet another storage-technology
available (PRAM? Racetrack?), that benefits from specific strategies.

So the competition will stay open... :-)

Regards,

Peter Niemayer


_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

^ permalink raw reply	[flat|nested] 11+ messages in thread

* Re: observed significant performance improvement using "delaylog" in
  2010-08-12 21:46   ` Peter Niemayer
@ 2010-08-13  9:56     ` Stan Hoeppner
  2010-08-13 10:35       ` Michael Monnerie
  0 siblings, 1 reply; 11+ messages in thread
From: Stan Hoeppner @ 2010-08-13  9:56 UTC (permalink / raw)
  To: xfs

Peter Niemayer put forth on 8/12/2010 4:46 PM:

> To me, as of today, XFS' big strength is performing good to
> excellent (while not always better than all other file-systems)
> in many use-cases - without worries about instability or immaturity.

+1  But don't forget the excellent tool set which I think is a little better
than those of the other filesystems.  XFS has the only functional/reliable
online defrag tool AFAIK.  The only 'missing' feature compared to the other
FSes is filesystem shrink ability, but for the majority of XFS users, I'm
guessing this isn't an issue as data sets are always growing.

Some benchmark results maybe worth a look:
http://btrfs.boxacle.net/repository/raid/2.6.35-rc5/2.6.35-rc5/

-- 
Stan

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

^ permalink raw reply	[flat|nested] 11+ messages in thread

* Re: observed significant performance improvement using "delaylog" in
  2010-08-13  9:56     ` Stan Hoeppner
@ 2010-08-13 10:35       ` Michael Monnerie
  2010-08-13 12:29         ` Christoph Hellwig
  0 siblings, 1 reply; 11+ messages in thread
From: Michael Monnerie @ 2010-08-13 10:35 UTC (permalink / raw)
  To: xfs; +Cc: Stan Hoeppner


[-- Attachment #1.1: Type: Text/Plain, Size: 916 bytes --]

On Freitag, 13. August 2010 Stan Hoeppner wrote:
> Some benchmark results maybe worth a look:
> http://btrfs.boxacle.net/repository/raid/2.6.35-rc5/2.6.35-rc5/
 
Thanks - it would have been great to see xfs with delaylog in that 
comparison, but the graphs are very very nice.

XFS seems performing better the more threads there are, just in "large 
file random reads" it's the slowest - why this?

A test with "lots of small files in a dir" would be great, something 
like the squid cachedir or just a single dir with >5.000 files in it.

-- 
mit freundlichen Grüssen,
Michael Monnerie, Ing. BSc

it-management Internet Services
http://proteger.at [gesprochen: Prot-e-schee]
Tel: 0660 / 415 65 31

****** Aktuelles Radiointerview! ******
http://www.it-podcast.at/aktuelle-sendung.html

// Wir haben im Moment zwei Häuser zu verkaufen:
// http://zmi.at/langegg/
// http://zmi.at/haus2009/

[-- Attachment #1.2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 198 bytes --]

[-- Attachment #2: Type: text/plain, Size: 121 bytes --]

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

^ permalink raw reply	[flat|nested] 11+ messages in thread

* Re: observed significant performance improvement using "delaylog" in
  2010-08-13 10:35       ` Michael Monnerie
@ 2010-08-13 12:29         ` Christoph Hellwig
  2010-08-13 14:13           ` Dave Chinner
  0 siblings, 1 reply; 11+ messages in thread
From: Christoph Hellwig @ 2010-08-13 12:29 UTC (permalink / raw)
  To: Michael Monnerie; +Cc: Stan Hoeppner, xfs

On Fri, Aug 13, 2010 at 12:35:44PM +0200, Michael Monnerie wrote:
> On Freitag, 13. August 2010 Stan Hoeppner wrote:
> > Some benchmark results maybe worth a look:
> > http://btrfs.boxacle.net/repository/raid/2.6.35-rc5/2.6.35-rc5/
>  
> Thanks - it would have been great to see xfs with delaylog in that 
> comparison, but the graphs are very very nice.
> 
> XFS seems performing better the more threads there are, just in "large 
> file random reads" it's the slowest - why this?

Any idea who is doing these runs?  Once we figure out what that large
file random reads loads is I'm sure we could fix it soon.  And asking
him/her to add -o delaylog would also be good.

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

^ permalink raw reply	[flat|nested] 11+ messages in thread

* Re: observed significant performance improvement using "delaylog" in
  2010-08-13 12:29         ` Christoph Hellwig
@ 2010-08-13 14:13           ` Dave Chinner
  2010-08-13 20:42             ` Eric Sandeen
  0 siblings, 1 reply; 11+ messages in thread
From: Dave Chinner @ 2010-08-13 14:13 UTC (permalink / raw)
  To: Christoph Hellwig; +Cc: Michael Monnerie, Stan Hoeppner, xfs

On Fri, Aug 13, 2010 at 08:29:07AM -0400, Christoph Hellwig wrote:
> On Fri, Aug 13, 2010 at 12:35:44PM +0200, Michael Monnerie wrote:
> > On Freitag, 13. August 2010 Stan Hoeppner wrote:
> > > Some benchmark results maybe worth a look:
> > > http://btrfs.boxacle.net/repository/raid/2.6.35-rc5/2.6.35-rc5/
> >  
> > Thanks - it would have been great to see xfs with delaylog in that 
> > comparison, but the graphs are very very nice.
> > 
> > XFS seems performing better the more threads there are, just in "large 
> > file random reads" it's the slowest - why this?
> 
> Any idea who is doing these runs?

IIRC the tests are run by someone from IBM, but I cannot remember
who it is.

> Once we figure out what that large
> file random reads loads is I'm sure we could fix it soon.

>From http://btrfs.boxacle.net/:

Random Reads (raid, single-disk) 
	Start with 1024 files. 
		100 MB files on the raid system.
		35 MB files on the single-disk system.
	Each thread reads a fixed amount of data from a random location in one file using 4 kB reads. 
		5 MB reads on the raid system.
		1 MB reads on the single-disk system.

So it's not a small random read workload (100GB data set), so the
files on XFS are probably more spread out over multiple AGs
and hence further apart than other filesystems. Hence a greater
average seek distance, hence it slower throughput....

> And asking
> him/her to add -o delaylog would also be good.

Yes, that would be an interesting comparison...

Cheers,

Dave.

-- 
Dave Chinner
david@fromorbit.com

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

^ permalink raw reply	[flat|nested] 11+ messages in thread

* Re: observed significant performance improvement using "delaylog" in
  2010-08-13 14:13           ` Dave Chinner
@ 2010-08-13 20:42             ` Eric Sandeen
  2010-08-14 11:28               ` Martin Steigerwald
  0 siblings, 1 reply; 11+ messages in thread
From: Eric Sandeen @ 2010-08-13 20:42 UTC (permalink / raw)
  To: Dave Chinner; +Cc: Christoph Hellwig, Michael Monnerie, Stan Hoeppner, xfs

Dave Chinner wrote:
> On Fri, Aug 13, 2010 at 08:29:07AM -0400, Christoph Hellwig wrote:

...

>> And asking
>> him/her to add -o delaylog would also be good.
> 
> Yes, that would be an interesting comparison...

I was thinking the same thing; I'll do that.

-Eric

> Cheers,
> 
> Dave.
> 

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

^ permalink raw reply	[flat|nested] 11+ messages in thread

* Re: observed significant performance improvement using "delaylog" in
  2010-08-13 20:42             ` Eric Sandeen
@ 2010-08-14 11:28               ` Martin Steigerwald
  2010-08-16  0:30                 ` Steven Pratt
  0 siblings, 1 reply; 11+ messages in thread
From: Martin Steigerwald @ 2010-08-14 11:28 UTC (permalink / raw)
  To: xfs
  Cc: Michael Monnerie, Eric Sandeen, Steven Pratt, Christoph Hellwig,
	Stan Hoeppner


Hi!

Cc'd Steve as I think he is running the benchmarks.

Am Freitag 13 August 2010 schrieb Eric Sandeen:
> Dave Chinner wrote:
> > On Fri, Aug 13, 2010 at 08:29:07AM -0400, Christoph Hellwig wrote:
> ...
> 
> >> And asking
> >> him/her to add -o delaylog would also be good.
> > 
> > Yes, that would be an interesting comparison...
> 
> I was thinking the same thing; I'll do that.

I think its Steven Pratt from IBM.

Steve, could you add a case for mount option "-o delaylog" to your 
benchmark runs?

----------  Weitergeleitete Nachricht  ----------

Subject: New performance results
Date: Samstag 24 April 2010
From: Steven Pratt <slpratt@austin.ibm.com>
To: "linux-btrfs" <linux-btrfs@vger.kernel.org>

Well, it's been a while but been very busy on other things. Thanks to 
Keith Mannthey for helping get some of these runs done.

New raid performance on 2.6.34-rc3:

http://btrfs.boxacle.net/repository/raid/2010-04-14_2004/2.6.34-
rc3/2.6.34-rc3.html

Also, the long running history graphs are here:

http://btrfs.boxacle.net/repository/raid/history/History.html


Steve


--
[...]
-------------------------------------------------------

Ciao,
-- 
Martin 'Helios' Steigerwald - http://www.Lichtvoll.de
GPG: 03B0 0D6C 0040 0710 4AFA  B82F 991B EAAC A599 84C7

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

^ permalink raw reply	[flat|nested] 11+ messages in thread

* Re: observed significant performance improvement using "delaylog" in
  2010-08-14 11:28               ` Martin Steigerwald
@ 2010-08-16  0:30                 ` Steven Pratt
  2010-08-16  1:03                   ` Dave Chinner
  0 siblings, 1 reply; 11+ messages in thread
From: Steven Pratt @ 2010-08-16  0:30 UTC (permalink / raw)
  To: Martin Steigerwald
  Cc: Michael Monnerie, Eric Sandeen, xfs, Christoph Hellwig,
	Keith Mannthey, Stan Hoeppner

Martin Steigerwald wrote:
> Hi!
>
> Cc'd Steve as I think he is running the benchmarks.
>
> Am Freitag 13 August 2010 schrieb Eric Sandeen:
>   
>> Dave Chinner wrote:
>>     
>>> On Fri, Aug 13, 2010 at 08:29:07AM -0400, Christoph Hellwig wrote:
>>>       
>> ...
>>
>>     
>>>> And asking
>>>> him/her to add -o delaylog would also be good.
>>>>         
>>> Yes, that would be an interesting comparison...
>>>       
>> I was thinking the same thing; I'll do that.
>>     
>
> I think its Steven Pratt from IBM.
>
> Steve, could you add a case for mount option "-o delaylog" to your 
> benchmark runs?
>   
Which filesystem is this for?  Keith has been helping me with some of 
these runs. We'll see if we can add this.

Steve

> ----------  Weitergeleitete Nachricht  ----------
>
> Subject: New performance results
> Date: Samstag 24 April 2010
> From: Steven Pratt <slpratt@austin.ibm.com>
> To: "linux-btrfs" <linux-btrfs@vger.kernel.org>
>
> Well, it's been a while but been very busy on other things. Thanks to 
> Keith Mannthey for helping get some of these runs done.
>
> New raid performance on 2.6.34-rc3:
>
> http://btrfs.boxacle.net/repository/raid/2010-04-14_2004/2.6.34-
> rc3/2.6.34-rc3.html
>
> Also, the long running history graphs are here:
>
> http://btrfs.boxacle.net/repository/raid/history/History.html
>
>
> Steve
>
>
> --
> [...]
> -------------------------------------------------------
>
> Ciao,
>   

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

^ permalink raw reply	[flat|nested] 11+ messages in thread

* Re: observed significant performance improvement using "delaylog" in
  2010-08-16  0:30                 ` Steven Pratt
@ 2010-08-16  1:03                   ` Dave Chinner
  0 siblings, 0 replies; 11+ messages in thread
From: Dave Chinner @ 2010-08-16  1:03 UTC (permalink / raw)
  To: Steven Pratt
  Cc: Michael Monnerie, Eric Sandeen, xfs, Christoph Hellwig,
	Keith Mannthey, Stan Hoeppner

On Sun, Aug 15, 2010 at 07:30:09PM -0500, Steven Pratt wrote:
> Martin Steigerwald wrote:
> >Hi!
> >
> >Cc'd Steve as I think he is running the benchmarks.
> >
> >Am Freitag 13 August 2010 schrieb Eric Sandeen:
> >>Dave Chinner wrote:
> >>>On Fri, Aug 13, 2010 at 08:29:07AM -0400, Christoph Hellwig wrote:
> >>...
> >>
> >>>>And asking
> >>>>him/her to add -o delaylog would also be good.
> >>>Yes, that would be an interesting comparison...
> >>I was thinking the same thing; I'll do that.
> >
> >I think its Steven Pratt from IBM.
> >
> >Steve, could you add a case for mount option "-o delaylog" to your
> >benchmark runs?
> Which filesystem is this for?  Keith has been helping me with some
> of these runs. We'll see if we can add this.

XFS - it's a new mount option for 2.6.35. Gory details in
Documentation/filesystems/xfs-delayed-logging-design.txt.

Cheers,

Dave.

-- 
Dave Chinner
david@fromorbit.com

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

^ permalink raw reply	[flat|nested] 11+ messages in thread

end of thread, other threads:[~2010-08-16  1:03 UTC | newest]

Thread overview: 11+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2010-08-12 10:46 observed significant performance improvement using "delaylog" in Khelben Blackstaff
2010-08-12 19:05 ` Michael Monnerie
2010-08-12 21:46   ` Peter Niemayer
2010-08-13  9:56     ` Stan Hoeppner
2010-08-13 10:35       ` Michael Monnerie
2010-08-13 12:29         ` Christoph Hellwig
2010-08-13 14:13           ` Dave Chinner
2010-08-13 20:42             ` Eric Sandeen
2010-08-14 11:28               ` Martin Steigerwald
2010-08-16  0:30                 ` Steven Pratt
2010-08-16  1:03                   ` Dave Chinner

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.