Responding to myself again, with more details...
On Fri, Oct 31, 2003 at 04:50:13PM -0500,
Andrew Klaassen wrote:
> I've been experiencing *very* slow sustained write ( < .5MB/sec)
> using 2.4.20-20.9.XFS1.3.1smp on a dual-Xeon box with a 3ware
> 7506 card in JBOD mode. Software RAID 5. No such problems with
> Ext3 ( > 20MB/sec sustained write) with the same kernel on the
> same software RAID 5 device.
I got those results using fs-bench:
http://h2np.net/tools/fs-bench.tar.gz
...using a little python script that simply monitors df -k.
Bonnie++ gives very different results; still favouring Ext3, but
the XFS results are at least acceptable:
Ext3:
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
thor.dkp.com 10000M 42713 44 33594 29 140724 49 215.5 1
XFS:
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
thor.dkp.com 10000M 20418 11 29688 16 134639 45 235.0 1
Hmm... so:
Would there either be 1) something about fs-bench that would
cause XFS to perform poorly or 2) something about watching df -k
with XFS that gives misleading results?
If anyone else could try fs-bench on an XFS-over-software-RAID5
setup and let me know their results, I'd be much obliged.
Andrew Klaassen
|