* Reiser4 / Reiserfs bonnie++ performance comparison
@ 2004-07-09 14:59 Philippe Gramoullé
2004-07-09 15:24 ` Marcel Hilzinger
` (2 more replies)
0 siblings, 3 replies; 12+ messages in thread
From: Philippe Gramoullé @ 2004-07-09 14:59 UTC (permalink / raw)
To: ReiserFS Mailing List
Hello,
Below are results from bonnie++ done on reiser4 and reiserfs.
Each result is an average of 5 runs.
For Reiser4 i used default mkfs options and -o formatting=extents
For Reisefs i used default mount options and -o notails
Assuming that default behavior of bonnie++ is to use twice the RAM of the server,
no reboot was done between consecutive runs
Bonnie++ was run like this : bonnie++ -x 5 -d bonnie
Hardware: Dell PowerEdge 2550 1GHz SMP, 1Go RAM, SCSI 15K RPM U160 Seagate Cheetah X15 Disk
Software: Debian Sid + 2004.07.02 Reiser4 auto snapshot ( 2.6.7-mm4 )
.config file available here if interested: http://philou.org/linux/reiser4/config-2.6.7-mm4
Biggest regression seems to be in deletes.
I will gladly redo my benchmarks if you feel i did or missed something obviously important,
or you'd like me to test other FSes as well
Comments are welcome.
Thanks,
Philippe
--
Lycos Europe Noc
Average of 5 runs with reiser4 ( no special mkfs options )
Version 1.03 ------Sequential Output------ --Sequential Input- --Random-
-Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP
localbox 2G 15357 93 29126 24 17855 24 15994 93 40185 26 322 1
------Sequential Create------ --------Random Create--------
-Create-- --Read--- -Delete-- -Create-- --Read--- -Delete--
files /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP
16 19839 99 +++++ +++ 7394 99.8 7632 100 +++++ +++ 7326 99.8
Average of 5 runs with reiser4 ( mkfs option -o formatting=extents)
Version 1.03 ------Sequential Output------ --Sequential Input- --Random-
-Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP
localbox 2G 15221 92 30467 25 17341 23 16059 94 40183 26 319 1
------Sequential Create------ --------Random Create--------
-Create-- --Read--- -Delete-- -Create-- --Read--- -Delete--
files /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP
16 19728 99 +++++ +++ 7361 99.8 7576 99 +++++ +++ 7265 99.8
Average of 5 runs with reiserfs ( no special mount options )
Version 1.03 ------Sequential Output------ --Sequential Input- --Random-
-Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP
localbox 2G 16207 97 30432 30 16978 16 15100 91 38487 22 349 1
------Sequential Create------ --------Random Create--------
-Create-- --Read--- -Delete-- -Create-- --Read--- -Delete--
files /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP
16 12433 99 +++++ +++ 10349 96 11836 99 +++++ +++ 9248 99.8
Average of 5 runs with reiserfs ( mount option -o notails)
Version 1.03 ------Sequential Output------ --Sequential Input- --Random-
-Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP
localbox 2G 16029 96 33199 33 16201 15 15226 91 38606 22 348 1
------Sequential Create------ --------Random Create--------
-Create-- --Read--- -Delete-- -Create-- --Read--- -Delete--
files /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP
16 12423 99 +++++ +++ 10719 99 11767 99 +++++ +++ 9241 99.4
^ permalink raw reply [flat|nested] 12+ messages in thread
* Re: Reiser4 / Reiserfs bonnie++ performance comparison
2004-07-09 14:59 Reiser4 / Reiserfs bonnie++ performance comparison Philippe Gramoullé
@ 2004-07-09 15:24 ` Marcel Hilzinger
2004-07-09 15:41 ` Chris Humphries
2004-07-09 18:37 ` Hans Reiser
2004-07-09 20:54 ` Redeeman
2 siblings, 1 reply; 12+ messages in thread
From: Marcel Hilzinger @ 2004-07-09 15:24 UTC (permalink / raw)
To: reiserfs-list
2004. július 9. 16.59 dátummal Philippe Gramoullé ezt írta:
> Hello,
>
> Below are results from bonnie++ done on reiser4 and reiserfs.
> Each result is an average of 5 runs.
>
> For Reiser4 i used default mkfs options and -o formatting=extents
> For Reisefs i used default mount options and -o notails
>
> Assuming that default behavior of bonnie++ is to use twice the RAM of the
> server, no reboot was done between consecutive runs
>
> Bonnie++ was run like this : bonnie++ -x 5 -d bonnie
>
> Hardware: Dell PowerEdge 2550 1GHz SMP, 1Go RAM, SCSI 15K RPM U160 Seagate
> Cheetah X15 Disk Software: Debian Sid + 2004.07.02 Reiser4 auto snapshot (
> 2.6.7-mm4 ) .config file available here if interested:
> http://philou.org/linux/reiser4/config-2.6.7-mm4
>
> Biggest regression seems to be in deletes.
You mustn't belive in Benchmarks :-). Try it empirically, and you will see,
that for big files, Reiser4 is much faster, then ReiserFS was. Some example
Deleting a 6,8GB DVD iso image (average of 10 tests):
ext3 14,15 sec
reiser3.6 12,67 sec
reiser4 1,23 sec
Deleting 3 GB of MP3 files (4MB average file size, 10 tests):
ext3 11,91 sec
reiser3.6 4,56 sec
reiser4 1,86 sec
Marcel
--
Üdvözlettel -- Mit freundlichen Grüssen,
Marcel Hilzinger
^ permalink raw reply [flat|nested] 12+ messages in thread
* Re: Reiser4 / Reiserfs bonnie++ performance comparison
2004-07-09 15:24 ` Marcel Hilzinger
@ 2004-07-09 15:41 ` Chris Humphries
2004-07-09 20:06 ` Marcelo Pacheco
2004-07-10 10:40 ` Mihai Rusu
0 siblings, 2 replies; 12+ messages in thread
From: Chris Humphries @ 2004-07-09 15:41 UTC (permalink / raw)
To: reiserfs-list
WHOA!
Marcel Hilzinger wrote:
>
> You mustn't belive in Benchmarks :-). Try it empirically, and you will see,
> that for big files, Reiser4 is much faster, then ReiserFS was. Some example
>
> Deleting a 6,8GB DVD iso image (average of 10 tests):
> ext3 14,15 sec
> reiser3.6 12,67 sec
> reiser4 1,23 sec
>
> Deleting 3 GB of MP3 files (4MB average file size, 10 tests):
> ext3 11,91 sec
> reiser3.6 4,56 sec
> reiser4 1,86 sec
>
> Marcel
^ permalink raw reply [flat|nested] 12+ messages in thread
* Re: Reiser4 / Reiserfs bonnie++ performance comparison
2004-07-09 14:59 Reiser4 / Reiserfs bonnie++ performance comparison Philippe Gramoullé
2004-07-09 15:24 ` Marcel Hilzinger
@ 2004-07-09 18:37 ` Hans Reiser
2004-07-12 14:58 ` Vladimir V. Saveliev
2004-07-09 20:54 ` Redeeman
2 siblings, 1 reply; 12+ messages in thread
From: Hans Reiser @ 2004-07-09 18:37 UTC (permalink / raw)
To: Philippe Gramoullé; +Cc: ReiserFS Mailing List, Vladimir Saveliev
Philippe Gramoulle' wrote:
> Hello,
>
>Below are results from bonnie++ done on reiser4 and reiserfs.
>Each result is an average of 5 runs.
>
>For Reiser4 i used default mkfs options and -o formatting=extents
>For Reisefs i used default mount options and -o notails
>
>Assuming that default behavior of bonnie++ is to use twice the RAM of the server,
>no reboot was done between consecutive runs
>
>Bonnie++ was run like this : bonnie++ -x 5 -d bonnie
>
>Hardware: Dell PowerEdge 2550 1GHz SMP, 1Go RAM, SCSI 15K RPM U160 Seagate Cheetah X15 Disk
>Software: Debian Sid + 2004.07.02 Reiser4 auto snapshot ( 2.6.7-mm4 )
>.config file available here if interested: http://philou.org/linux/reiser4/config-2.6.7-mm4
>
>Biggest regression seems to be in deletes.
>
>
I would like to know exactly what is being done in sequential block
output phase of this benchmark. vs, please comment on that and why
reiser4 is slow at it.
>I will gladly redo my benchmarks if you feel i did or missed something obviously important,
>or you'd like me to test other FSes as well
>
>Comments are welcome.
>
>Thanks,
>
>Philippe
>
>--
>Lycos Europe Noc
>
>
>
>Average of 5 runs with reiser4 ( no special mkfs options )
>
>Version 1.03 ------Sequential Output------ --Sequential Input- --Random-
> -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
>Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP
>localbox 2G 15357 93 29126 24 17855 24 15994 93 40185 26 322 1
> ------Sequential Create------ --------Random Create--------
> -Create-- --Read--- -Delete-- -Create-- --Read--- -Delete--
> files /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP
> 16 19839 99 +++++ +++ 7394 99.8 7632 100 +++++ +++ 7326 99.8
>
>
>Average of 5 runs with reiser4 ( mkfs option -o formatting=extents)
>
>Version 1.03 ------Sequential Output------ --Sequential Input- --Random-
> -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
>Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP
>localbox 2G 15221 92 30467 25 17341 23 16059 94 40183 26 319 1
> ------Sequential Create------ --------Random Create--------
> -Create-- --Read--- -Delete-- -Create-- --Read--- -Delete--
> files /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP
> 16 19728 99 +++++ +++ 7361 99.8 7576 99 +++++ +++ 7265 99.8
>
>
>Average of 5 runs with reiserfs ( no special mount options )
>
>Version 1.03 ------Sequential Output------ --Sequential Input- --Random-
> -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
>Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP
>localbox 2G 16207 97 30432 30 16978 16 15100 91 38487 22 349 1
> ------Sequential Create------ --------Random Create--------
> -Create-- --Read--- -Delete-- -Create-- --Read--- -Delete--
> files /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP
> 16 12433 99 +++++ +++ 10349 96 11836 99 +++++ +++ 9248 99.8
>
>
>Average of 5 runs with reiserfs ( mount option -o notails)
>
>Version 1.03 ------Sequential Output------ --Sequential Input- --Random-
> -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
>Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP
>localbox 2G 16029 96 33199 33 16201 15 15226 91 38606 22 348 1
> ------Sequential Create------ --------Random Create--------
> -Create-- --Read--- -Delete-- -Create-- --Read--- -Delete--
> files /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP
> 16 12423 99 +++++ +++ 10719 99 11767 99 +++++ +++ 9241 99.4
>
>
>
>
>
^ permalink raw reply [flat|nested] 12+ messages in thread
* Re: Reiser4 / Reiserfs bonnie++ performance comparison
2004-07-09 15:41 ` Chris Humphries
@ 2004-07-09 20:06 ` Marcelo Pacheco
2004-07-10 5:08 ` Hans Reiser
2004-07-10 10:40 ` Mihai Rusu
1 sibling, 1 reply; 12+ messages in thread
From: Marcelo Pacheco @ 2004-07-09 20:06 UTC (permalink / raw)
To: reiserfs-list
When you delete a file what you need to do is to remove the directory entry,
the inode and the file allocation structures. Reiser3 uses one key per file
block (4kb), while reiser4 uses extents, which could be huge (100's of MB)
per key, so there's far less things to remove when you remove a huge file on
Reiser4 than on Reiser3.
Marcelo Pacheco
Em Sex 09 Jul 2004 12:41, Chris Humphries escreveu:
> WHOA!
>
> Marcel Hilzinger wrote:
> > You mustn't belive in Benchmarks :-). Try it empirically, and you will
> > see, that for big files, Reiser4 is much faster, then ReiserFS was. Some
> > example
> >
> > Deleting a 6,8GB DVD iso image (average of 10 tests):
> > ext3 14,15 sec
> > reiser3.6 12,67 sec
> > reiser4 1,23 sec
> >
> > Deleting 3 GB of MP3 files (4MB average file size, 10 tests):
> > ext3 11,91 sec
> > reiser3.6 4,56 sec
> > reiser4 1,86 sec
> >
> > Marcel
^ permalink raw reply [flat|nested] 12+ messages in thread
* Re: Reiser4 / Reiserfs bonnie++ performance comparison
2004-07-09 14:59 Reiser4 / Reiserfs bonnie++ performance comparison Philippe Gramoullé
2004-07-09 15:24 ` Marcel Hilzinger
2004-07-09 18:37 ` Hans Reiser
@ 2004-07-09 20:54 ` Redeeman
2004-07-09 21:52 ` Dieter Nützel
2004-07-10 5:17 ` Hans Reiser
2 siblings, 2 replies; 12+ messages in thread
From: Redeeman @ 2004-07-09 20:54 UTC (permalink / raw)
To: Reiserfs Mailinglist
My benchmarks.. (if you can call it that)
reiser4
root@redeeman: /mnt/fstest# time tar -xjf /miscbinds/work/linux-
2.6.5.tar.bz2
real 0m18.834s
user 0m14.785s
sys 0m2.516s
root@redeeman: /mnt/fstest# time cp -a linux-2.6.5 linux
real 0m12.936s
user 0m0.067s
sys 0m2.137s
root@redeeman: /mnt/fstest# time rm -fr linux
real 0m3.596s
user 0m0.012s
sys 0m1.237s
-------------------------------------
reiserfs:
root@redeeman: /mnt/fstest# time tar -xjf /miscbinds/work/linux-
2.6.5.tar.bz2
real 0m31.078s
user 0m15.232s
sys 0m2.099s
root@redeeman: /mnt/fstest# time cp -a linux-2.6.5 linux
real 0m50.763s
user 0m0.134s
sys 0m2.935s
root@redeeman: /mnt/fstest# time rm -fr linux
real 0m1.493s
user 0m0.013s
sys 0m0.722s
---------------------------------------
xfs
root@redeeman: /mnt/fstest# time tar -xjf /miscbinds/work/linux-
2.6.5.tar.bz2
real 0m36.432s
user 0m15.202s
sys 0m1.727s
root@redeeman: /mnt/fstest# time cp -a linux-2.6.5 linux
real 1m10.314s
user 0m0.162s
sys 0m3.428s
root@redeeman: /mnt/fstest# time rm -fr linux
real 0m22.153s
user 0m0.042s
sys 0m1.618s
------------------------------------------------------------------------------
ext3
root@redeeman: /mnt/fstest# time tar -xjf /miscbinds/work/linux-
2.6.5.tar.bz2
real 0m32.082s
user 0m15.190s
sys 0m1.519s
root@redeeman: /mnt/fstest# time cp -a linux-2.6.5 linux
real 0m39.479s
user 0m0.095s
sys 0m2.059s
root@redeeman: /mnt/fstest# time rm -fr linux
real 0m0.846s
user 0m0.012s
sys 0m0.369s
--------------------------------------------------------------------------------
ext2
root@redeeman: /mnt/fstest# time tar -xjf /miscbinds/work/linux-
2.6.5.tar.bz2
real 0m30.046s
user 0m15.165s
sys 0m1.025s
root@redeeman: /mnt/fstest# time cp -a linux-2.6.5 linux
real 0m18.344s
user 0m0.104s
sys 0m1.454s
root@redeeman: /mnt/fstest# time rm -fr linux
real 0m1.367s
user 0m0.009s
sys 0m0.226s
---------------------------------------------------------------------------------
fat32
root@redeeman: /mnt/fstest# time tar -xjf /miscbinds/work/linux-
2.6.5.tar.bz2
real 0m45.634s
user 0m25.625s
sys 0m6.361s
root@redeeman: /mnt/fstest# time cp -a linux-2.6.5 linux
real 0m59.151s
user 0m9.098s
sys 0m17.761s
root@redeeman: /mnt/fstest# time rm -fr linux
real 0m2.449s
user 0m0.009s
sys 0m1.121s
On Fri, 2004-07-09 at 16:59 +0200, Philippe Gramoullé wrote:
> Hello,
>
> Below are results from bonnie++ done on reiser4 and reiserfs.
> Each result is an average of 5 runs.
>
> For Reiser4 i used default mkfs options and -o formatting=extents
> For Reisefs i used default mount options and -o notails
>
> Assuming that default behavior of bonnie++ is to use twice the RAM of the server,
> no reboot was done between consecutive runs
>
> Bonnie++ was run like this : bonnie++ -x 5 -d bonnie
>
> Hardware: Dell PowerEdge 2550 1GHz SMP, 1Go RAM, SCSI 15K RPM U160 Seagate Cheetah X15 Disk
> Software: Debian Sid + 2004.07.02 Reiser4 auto snapshot ( 2.6.7-mm4 )
> .config file available here if interested: http://philou.org/linux/reiser4/config-2.6.7-mm4
>
> Biggest regression seems to be in deletes.
>
> I will gladly redo my benchmarks if you feel i did or missed something obviously important,
> or you'd like me to test other FSes as well
>
> Comments are welcome.
>
> Thanks,
>
> Philippe
>
> --
> Lycos Europe Noc
>
>
>
> Average of 5 runs with reiser4 ( no special mkfs options )
>
> Version 1.03 ------Sequential Output------ --Sequential Input- --Random-
> -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
> Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP
> localbox 2G 15357 93 29126 24 17855 24 15994 93 40185 26 322 1
> ------Sequential Create------ --------Random Create--------
> -Create-- --Read--- -Delete-- -Create-- --Read--- -Delete--
> files /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP
> 16 19839 99 +++++ +++ 7394 99.8 7632 100 +++++ +++ 7326 99.8
>
>
> Average of 5 runs with reiser4 ( mkfs option -o formatting=extents)
>
> Version 1.03 ------Sequential Output------ --Sequential Input- --Random-
> -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
> Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP
> localbox 2G 15221 92 30467 25 17341 23 16059 94 40183 26 319 1
> ------Sequential Create------ --------Random Create--------
> -Create-- --Read--- -Delete-- -Create-- --Read--- -Delete--
> files /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP
> 16 19728 99 +++++ +++ 7361 99.8 7576 99 +++++ +++ 7265 99.8
>
>
> Average of 5 runs with reiserfs ( no special mount options )
>
> Version 1.03 ------Sequential Output------ --Sequential Input- --Random-
> -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
> Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP
> localbox 2G 16207 97 30432 30 16978 16 15100 91 38487 22 349 1
> ------Sequential Create------ --------Random Create--------
> -Create-- --Read--- -Delete-- -Create-- --Read--- -Delete--
> files /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP
> 16 12433 99 +++++ +++ 10349 96 11836 99 +++++ +++ 9248 99.8
>
>
> Average of 5 runs with reiserfs ( mount option -o notails)
>
> Version 1.03 ------Sequential Output------ --Sequential Input- --Random-
> -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
> Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP
> localbox 2G 16029 96 33199 33 16201 15 15226 91 38606 22 348 1
> ------Sequential Create------ --------Random Create--------
> -Create-- --Read--- -Delete-- -Create-- --Read--- -Delete--
> files /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP
> 16 12423 99 +++++ +++ 10719 99 11767 99 +++++ +++ 9241 99.4
>
>
--
Redeeman <redeeman@metanurb.dk>
^ permalink raw reply [flat|nested] 12+ messages in thread
* Re: Reiser4 / Reiserfs bonnie++ performance comparison
2004-07-09 20:54 ` Redeeman
@ 2004-07-09 21:52 ` Dieter Nützel
2004-07-09 22:06 ` Redeeman
2004-07-10 5:17 ` Hans Reiser
1 sibling, 1 reply; 12+ messages in thread
From: Dieter Nützel @ 2004-07-09 21:52 UTC (permalink / raw)
To: reiserfs-list, redeeman
Am Freitag, 9. Juli 2004 22:54 schrieb Redeeman:
> My benchmarks.. (if you can call it that)
> reiser4
>
> root@redeeman: /mnt/fstest# time tar -xjf /miscbinds/work/linux-
> 2.6.5.tar.bz2
>
> real 0m18.834s
> user 0m14.785s
> sys 0m2.516s
>
> root@redeeman: /mnt/fstest# time cp -a linux-2.6.5 linux
>
> real 0m12.936s
> user 0m0.067s
> sys 0m2.137s
>
> root@redeeman: /mnt/fstest# time rm -fr linux
>
> real 0m3.596s
> user 0m0.012s
> sys 0m1.237s
>
> -------------------------------------
>
> reiserfs:
>
> root@redeeman: /mnt/fstest# time tar -xjf /miscbinds/work/linux-
> 2.6.5.tar.bz2
>
> real 0m31.078s
> user 0m15.232s
> sys 0m2.099s
>
> root@redeeman: /mnt/fstest# time cp -a linux-2.6.5 linux
>
> real 0m50.763s
> user 0m0.134s
> sys 0m2.935s
>
> root@redeeman: /mnt/fstest# time rm -fr linux
>
> real 0m1.493s
> user 0m0.013s
> sys 0m0.722s
All ReiserFS 3.6 without data-logging (parallel write optimization, etc.), NEW
block allocation, with tails?
2.6.7-mmX (or Chris patches)?
Thanks,
Dieter
^ permalink raw reply [flat|nested] 12+ messages in thread
* Re: Reiser4 / Reiserfs bonnie++ performance comparison
2004-07-09 21:52 ` Dieter Nützel
@ 2004-07-09 22:06 ` Redeeman
0 siblings, 0 replies; 12+ messages in thread
From: Redeeman @ 2004-07-09 22:06 UTC (permalink / raw)
To: Reiserfs Mailinglist
this was done using the 2.6.5-rc2 stable snapshot, and the stock
reiserfs in the kernel
On Fri, 2004-07-09 at 23:52 +0200, Dieter Nützel wrote:
> Am Freitag, 9. Juli 2004 22:54 schrieb Redeeman:
> > My benchmarks.. (if you can call it that)
> > reiser4
> >
> > root@redeeman: /mnt/fstest# time tar -xjf /miscbinds/work/linux-
> > 2.6.5.tar.bz2
> >
> > real 0m18.834s
> > user 0m14.785s
> > sys 0m2.516s
> >
> > root@redeeman: /mnt/fstest# time cp -a linux-2.6.5 linux
> >
> > real 0m12.936s
> > user 0m0.067s
> > sys 0m2.137s
> >
> > root@redeeman: /mnt/fstest# time rm -fr linux
> >
> > real 0m3.596s
> > user 0m0.012s
> > sys 0m1.237s
> >
> > -------------------------------------
> >
> > reiserfs:
> >
> > root@redeeman: /mnt/fstest# time tar -xjf /miscbinds/work/linux-
> > 2.6.5.tar.bz2
> >
> > real 0m31.078s
> > user 0m15.232s
> > sys 0m2.099s
> >
> > root@redeeman: /mnt/fstest# time cp -a linux-2.6.5 linux
> >
> > real 0m50.763s
> > user 0m0.134s
> > sys 0m2.935s
> >
> > root@redeeman: /mnt/fstest# time rm -fr linux
> >
> > real 0m1.493s
> > user 0m0.013s
> > sys 0m0.722s
>
> All ReiserFS 3.6 without data-logging (parallel write optimization, etc.), NEW
> block allocation, with tails?
>
> 2.6.7-mmX (or Chris patches)?
>
> Thanks,
> Dieter
>
--
Redeeman <redeeman@metanurb.dk>
^ permalink raw reply [flat|nested] 12+ messages in thread
* Re: Reiser4 / Reiserfs bonnie++ performance comparison
2004-07-09 20:06 ` Marcelo Pacheco
@ 2004-07-10 5:08 ` Hans Reiser
0 siblings, 0 replies; 12+ messages in thread
From: Hans Reiser @ 2004-07-10 5:08 UTC (permalink / raw)
To: Marcelo Pacheco; +Cc: reiserfs-list
Marcelo Pacheco wrote:
>When you delete a file what you need to do is to remove the directory entry,
>the inode and the file allocation structures. Reiser3 uses one key
>
^key^pointer
> per file
>block (4kb), while reiser4 uses extents, which could be huge (100's of MB)
>per key,
>
^key^pointer
> so there's far less things to remove when you remove a huge file on
>Reiser4 than on Reiser3.
>
>Marcelo Pacheco
>
>Em Sex 09 Jul 2004 12:41, Chris Humphries escreveu:
>
>
>>WHOA!
>>
>>Marcel Hilzinger wrote:
>>
>>
>>>You mustn't belive in Benchmarks :-). Try it empirically, and you will
>>>see, that for big files, Reiser4 is much faster, then ReiserFS was. Some
>>>example
>>>
>>>Deleting a 6,8GB DVD iso image (average of 10 tests):
>>>ext3 14,15 sec
>>>reiser3.6 12,67 sec
>>>reiser4 1,23 sec
>>>
>>>Deleting 3 GB of MP3 files (4MB average file size, 10 tests):
>>>ext3 11,91 sec
>>>reiser3.6 4,56 sec
>>>reiser4 1,86 sec
>>>
>>>Marcel
>>>
>>>
>
>
>
>
^ permalink raw reply [flat|nested] 12+ messages in thread
* Re: Reiser4 / Reiserfs bonnie++ performance comparison
2004-07-09 20:54 ` Redeeman
2004-07-09 21:52 ` Dieter Nützel
@ 2004-07-10 5:17 ` Hans Reiser
1 sibling, 0 replies; 12+ messages in thread
From: Hans Reiser @ 2004-07-10 5:17 UTC (permalink / raw)
To: redeeman; +Cc: Reiserfs Mailinglist
So one of the things that does not make sense to me in these benchmarks
is that reiser4 with extents only is much slower than reiserfs V3 with
notails at deletes.
Ideas anyone?
Is it maybe because V3 puts the directory entries for multiple
directories near each other? What happens when you delete one huge
directory instead of a tree?
^ permalink raw reply [flat|nested] 12+ messages in thread
* Re: Reiser4 / Reiserfs bonnie++ performance comparison
2004-07-09 15:41 ` Chris Humphries
2004-07-09 20:06 ` Marcelo Pacheco
@ 2004-07-10 10:40 ` Mihai Rusu
1 sibling, 0 replies; 12+ messages in thread
From: Mihai Rusu @ 2004-07-10 10:40 UTC (permalink / raw)
To: Chris Humphries; +Cc: reiserfs-list
On Fri, 9 Jul 2004, Chris Humphries wrote:
> WHOA!
Try benchmarking XFS for deleting large files too if there is any point in
that :)
Anyway, bonnie++ offers a very good benchmark for those interested in
having large mail servers, web servers with tons of small filess
created/read/deleted randomly. And that is where reiserfs3/4 shines above
all others except the reiser4 regression with its random deletes which I
also reported it some time ago on this list :-/
>
> Marcel Hilzinger wrote:
> >
> > You mustn't belive in Benchmarks :-). Try it empirically, and you will see,
> > that for big files, Reiser4 is much faster, then ReiserFS was. Some example
> >
> > Deleting a 6,8GB DVD iso image (average of 10 tests):
> > ext3 14,15 sec
> > reiser3.6 12,67 sec
> > reiser4 1,23 sec
> >
> > Deleting 3 GB of MP3 files (4MB average file size, 10 tests):
> > ext3 11,91 sec
> > reiser3.6 4,56 sec
> > reiser4 1,86 sec
> >
> > Marcel
>
>
--
Mihai RUSU Email: dizzy@roedu.net
GPG : http://dizzy.roedu.net/dizzy-gpg.txt WWW: http://dizzy.roedu.net
"Linux is obsolete" -- AST
^ permalink raw reply [flat|nested] 12+ messages in thread
* Re: Reiser4 / Reiserfs bonnie++ performance comparison
2004-07-09 18:37 ` Hans Reiser
@ 2004-07-12 14:58 ` Vladimir V. Saveliev
0 siblings, 0 replies; 12+ messages in thread
From: Vladimir V. Saveliev @ 2004-07-12 14:58 UTC (permalink / raw)
To: Hans Reiser; +Cc: ReiserFS Mailing List
Hans Reiser wrote:
> Philippe Gramoulle' wrote:
>
>> Hello,
>>
>> Below are results from bonnie++ done on reiser4 and reiserfs.
>> Each result is an average of 5 runs.
>>
>> For Reiser4 i used default mkfs options and -o formatting=extents
>> For Reisefs i used default mount options and -o notails
>>
>> Assuming that default behavior of bonnie++ is to use twice the RAM of
>> the server,
>> no reboot was done between consecutive runs
>>
>> Bonnie++ was run like this : bonnie++ -x 5 -d bonnie
>>
>> Hardware: Dell PowerEdge 2550 1GHz SMP, 1Go RAM, SCSI 15K RPM U160
>> Seagate Cheetah X15 Disk
>> Software: Debian Sid + 2004.07.02 Reiser4 auto snapshot ( 2.6.7-mm4 )
>> .config file available here if interested:
>> http://philou.org/linux/reiser4/config-2.6.7-mm4
>>
>> Biggest regression seems to be in deletes.
>>
>>
> I would like to know exactly what is being done in sequential block
> output phase of this benchmark. vs, please comment on that and why
> reiser4 is slow at it.
>
sequential block output of bonnie++ writes to a file by 8192-byte blocks with write(2).
Reiser4 does it similar to reiserfs which is known as fast filesystem. In my tests I get the following.
As you can see Per Char output is worse in reiser4 than in reiserfs.
Version 1.03 ------Sequential Output------ --Sequential Input- --Random-
-Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP
reiser4 (1) 1G 18408 81 29188 15 13575 14 18317 78 30200 9 257.1 1
(2) 1G 17386 76 27335 15 13434 12 18617 79 30139 9 239.9 1
(3) 1G 18633 82 25161 13 14119 12 18292 78 30178 9 236.6 1
(4) 1G 17318 76 27457 15 12753 12 19083 81 29813 9 237.3 1
(5) 1G 19211 84 25424 13 13865 12 17785 76 29839 8 245.9 1
-------------------------------------------------------------------------------
reiserfs (1) 1G 21027 93 25964 23 10808 5 14502 62 23672 6 245.6 0
(2) 1G 22128 97 25173 21 10779 5 14354 62 23708 6 251.9 0
(3) 1G 21154 93 25794 22 10638 5 14513 62 23676 6 248.9 0
>> I will gladly redo my benchmarks if you feel i did or missed something
>> obviously important,
>> or you'd like me to test other FSes as well
>>
>> Comments are welcome.
>>
>> Thanks,
>>
>> Philippe
>>
>> --
>> Lycos Europe Noc
>>
>>
>>
>> Average of 5 runs with reiser4 ( no special mkfs options )
>>
>> Version 1.03 ------Sequential Output------ --Sequential Input-
>> --Random-
>> -Per Chr- --Block-- -Rewrite- -Per Chr- --Block--
>> --Seeks--
>> Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP
>> /sec %CP
>> localbox 2G 15357 93 29126 24 17855 24 15994 93
>> 40185 26 322 1
>> ------Sequential Create------ --------Random
>> Create--------
>> -Create-- --Read--- -Delete-- -Create-- --Read---
>> -Delete--
>> files /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP
>> /sec %CP
>> 16 19839 99 +++++ +++ 7394 99.8 7632 100 +++++
>> +++ 7326 99.8
>>
>>
>> Average of 5 runs with reiser4 ( mkfs option -o formatting=extents)
>>
>> Version 1.03 ------Sequential Output------ --Sequential Input-
>> --Random-
>> -Per Chr- --Block-- -Rewrite- -Per Chr- --Block--
>> --Seeks--
>> Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP
>> /sec %CP
>> localbox 2G 15221 92 30467 25 17341 23 16059 94 40183
>> 26 319 1
>> ------Sequential Create------ --------Random
>> Create--------
>> -Create-- --Read--- -Delete-- -Create-- --Read---
>> -Delete--
>> files /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP
>> /sec %CP
>> 16 19728 99 +++++ +++ 7361 99.8 7576 99 +++++
>> +++ 7265 99.8
>>
>> Average of 5 runs with reiserfs ( no special mount options )
>>
>> Version 1.03 ------Sequential Output------ --Sequential Input-
>> --Random-
>> -Per Chr- --Block-- -Rewrite- -Per Chr- --Block--
>> --Seeks--
>> Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP
>> /sec %CP
>> localbox 2G 16207 97 30432 30 16978 16 15100 91 38487
>> 22 349 1
>> ------Sequential Create------ --------Random
>> Create--------
>> -Create-- --Read--- -Delete-- -Create-- --Read---
>> -Delete--
>> files /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP
>> /sec %CP
>> 16 12433 99 +++++ +++ 10349 96 11836 99 +++++
>> +++ 9248 99.8
>> Average of 5 runs with reiserfs ( mount option -o
>> notails)
>>
>> Version 1.03 ------Sequential Output------ --Sequential Input-
>> --Random-
>> -Per Chr- --Block-- -Rewrite- -Per Chr- --Block--
>> --Seeks--
>> Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP
>> /sec %CP
>> localbox 2G 16029 96 33199 33 16201 15 15226 91 38606
>> 22 348 1
>> ------Sequential Create------ --------Random
>> Create--------
>> -Create-- --Read--- -Delete-- -Create-- --Read---
>> -Delete--
>> files /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP
>> /sec %CP
>> 16 12423 99 +++++ +++ 10719 99 11767 99 +++++
>> +++ 9241 99.4
>>
>>
>>
>>
>>
>
>
>
^ permalink raw reply [flat|nested] 12+ messages in thread
end of thread, other threads:[~2004-07-12 14:58 UTC | newest]
Thread overview: 12+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2004-07-09 14:59 Reiser4 / Reiserfs bonnie++ performance comparison Philippe Gramoullé
2004-07-09 15:24 ` Marcel Hilzinger
2004-07-09 15:41 ` Chris Humphries
2004-07-09 20:06 ` Marcelo Pacheco
2004-07-10 5:08 ` Hans Reiser
2004-07-10 10:40 ` Mihai Rusu
2004-07-09 18:37 ` Hans Reiser
2004-07-12 14:58 ` Vladimir V. Saveliev
2004-07-09 20:54 ` Redeeman
2004-07-09 21:52 ` Dieter Nützel
2004-07-09 22:06 ` Redeeman
2004-07-10 5:17 ` Hans Reiser
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.