| Author |
Message |
heavensrevenge
|
|
Post subject: Filesystem benckmark shootout
 Posted: Aug 13, 2007 - 01:10 AM
|
|
Joined: Jun 06, 2007
Posts: 27
Location: Canada
Status: Offline
|
|
Intro:
For write performance, the source drive I used was a 750 GB FreeAgent Pro external drive attached by USB 2.0 so the only thing the internal destination drive had to worry about was writing.
For the copy I used the same file system, mkfs options, and mount options to measure copy performance from 1 partition on the drive to another partition on the same drive.
The read performance was calculated by using the fs to fs copy minus write performance (Ex. xfs to xfs copy-xfs write=xfs read).
And the delete performance was simple, deleting and timing it.
All times were taken/recorded using the "stopwatch" application that is written in tk/tcl, all calculations were done using the amazing Qalculate QT calculator in time calculation format, and all file system transactions were done using Krusader(root mode).
All testing was done on the same box(exact same hardware) that utilized the sidux distro fully dist-upgraded. Running kernel 2.6.22.2.
reiser3 used the notail mount option
ext3 was using the noatime mount option and was journal_data_writeback enabled
xfs was using the nodiratime,noatime,logbufs=8,osyncisdsync,biosize=16 fstab mount options
All times are in minutes unless otherwise stated.
Benchmark:
Programming folder (variety of large+small files):
jfs write= 3:30
jfs to jfs copy= 4:49
jfs read= 1:19
jfs delete= 6s
reiser3 write= 3:13
reiser3 to reiser3 copy= 4:10
reiser3 read= 57s
reiser3 delete= 6s
xfs write= 3:45
xfs to xfs copy= 5:02
xfs read= 1:17
xfs delete= 48s
ext3 write= 3:30
ext3 to ext3 copy= 4:28
ext3 read= 58s
ext3 delete= 6s
GNAT 2007 (8000 small files):
jfs write= 1:36
jfs to jfs copy= 1:50
jfs read= 14s
jfs delete= 3s
reiser3 write= 1:10
reiser3 to reiser3 copy= 1:20
reiser3 read= 10s
reiser3 delete= 1s
xfs write= 1:20
xfs to xfs copy= 1:46
xfs read= 26s
xfs delete= 30s
ext3 write= 1:18
ext3 to ext3 copy= 1:34
ext3 read= 16s
ext3 delete= 2
Dark Tranquility (~250 mp3's):
jfs write= 1:07
jfs to jfs copy= 2:02
jfs read= 55s
jfs delete= 2s
reiser3 write= 59s
reiser3 to reiser3 copy= 1:40
reiser3 read= 41s
reiser3 delete= 4s
xfs write= 1:02
xfs to xfs copy= 1:24
xfs read= 22s
xfs delete= 2s
ext3 write= 1:02
ext3 to ext3 copy= 1:52
ext3 read= 50s
ext3 delete= 4s
6 Distro CD ISO's (~700MB each):
jfs write= 2:24
jfs to jfs copy= 4:19
jfs read= 1:55
jfs delete= instant
reiser3 write= 2:28
reiser3 to reiser3 copy= 4:02
reiser3 read= 1:34
reiser3 delete= 6s
xfs write= 2:18
xfs to xfs copy= 3:13
xfs read= 55s
xfs delete= instant
ext3 write= 2:34
ext3 to ext3 copy= 3:48
ext3 read= 1:14
ext3 delete= 6s
12 Futurama from ext3 same hd (175MB each) :
jfs write= 1:59
jfs to jfs copy= 2:18
jfs read= 19s
jfs delete= instant
reiser3 write= 1:47
reiser3 to reiser3 copy= 2:00
reiser3 read= 13s
reiser3 delete= 2s
xfs write= 1:31
xfs to xfs copy= 1:34
xfs read= 3s
xfs delete= instant
ext3 write= 1:53
ext3 to ext3 copy= 1:56
ext3 read= 3s
ext3 delete= 3s
2 DVD ISO'S (7 GB total):
jfs write= 4:24
jfs to jfs copy= 7:38
jfs read= 3:14
jfs delete= instant
reiser3 write= 4:20
reiser3 to reiser3 copy= 7:36
reiser3 read= 3:16
reiser3 delete= 2s
xfs write= 4:16
xfs to xfs copy= 5:55
xfs read= 1:39
xfs delete= instant
ext3 write= 4:33
ext3 to ext3 copy= 7:00
ext3 read= 2:27
ext3 delete= 12s
TOTALS:
jfs write= 15 mins
jfs to jfs copy= 22:56 mins
jfs read= 7:56 mins
jfs delete= 11 seconds
jfs total time= 46:03
reiser write= 13:57 mins
reiser3 to reiser3 copy= 20:48 mins
reiser3 read= 6:51 mins
reiser3 delete= 19 seconds
reiser total time= 41:55
xfs write= 14:12 mins
xfs to xfs copy= 18:54 mins
xfs read= 4:42 mins
xfs delete= 1:20 mins
xfs total time= 39:08
ext3 write= 14:50 mins
ext3 to ext3 copy= 20:38 mins
ext3 read= 5:48 mins
ext3 delete= 33 seconds
ext3 total time= 41:49
4th place goes to jfs
3rd place goes to reiser3
2nd place goes to ext3
1st place goes to xfs
*+*+*+*+*+*+*+*+*+*+*+*+*+*+*+*+*+*+*+*
*+*+*+*+* ------ XFS WINS !!!!!! ------ *+*+*+*+*
*+*+*+*+*+*+*+*+*+*+*+*+*+*+*+*+*+*+*+*
Feel free to correct and suggest, constructive criticism is welcome.
I would have tested ext4 and reiser4 as well, but I was just using the stock sidux kernel, nothing custom/fancy quite yet, and also helps to address availability and what you'd currently like to choose right NOW.
-- HeavensRevenge |
Last edited by heavensrevenge on Aug 13, 2007 - 04:31 AM; edited 1 time in total
|
| |
|
|
|
 |
|
shame
|
|
|
Post subject:
Posted: Aug 13, 2007 - 01:45 AM
|
|
Joined: Dec 04, 2006
Posts: 1628
Location: England
Status: Offline
|
|
I've used various different filesystems and never noticed any difference, except reiser seemed to make everything slow after a few weeks.
I just use ext3 all the time now.
Maybe it just depends what you use the comp for... |
_________________
sidux 32 & 64 | Parsix | Mandriva Cooker
repository of shame - compiz fusion repo for sidux/debian | 32-bit | 64-bit
|
| |
|
|
|
|
|
slh
|
|
|
Post subject:
Posted: Aug 13, 2007 - 08:00 AM
|
|
Joined: Nov 25, 2006
Posts: 3216
Status: Offline
|
|
I wonder that xfs lasted long enough to survive your "benchmarks" (yes, this is slightly exaggerated, but only slightly).
Regarding the "benchmark" itself, unfortunately it doesn't really stand scientific methods, as you're USB/ and converter limited, not file system limited and don't go into details regarding the test setup, especially the type of data and the host system are left in the dark. |
|
|
| |
|
|
|
|
|
slam
|
|
|
Post subject:
Posted: Aug 13, 2007 - 09:01 AM
|
|
Team Member
Joined: Nov 24, 2006
w3
Posts: 3690
Status: Offline
|
|
Everybody is free to use whatever filesystem he has fun to test. However, we at sidux default to/strongly suggest/only support/ ext3 for various reasons which you might want to read up here.
Greetings,
Chris |
_________________
an operating system must operate - sidux.com
development is life - code.zikula.org
my sidux repo - w3you.com
|
| |
|
|
|
|
|
|
|
|
