John H Terpstra wrote:
Yes, I noticed.Phil,One of the things that really spooked me while doing my benchmark tests is the impact made by the file system type. I forgot to make mention of that in my reply. Ext2fs is by far the fastest file system on Linux. Ext3fs is the slowest, ReiserFS is in between them. One of the things I want to do soon is to benchmark XFS and JFS against Ext2fs and Ext3fs.
I use ext3 because a fsck on a 200GB+ ext2 filesystem kinda kills the productivity of one's co-workers - especially sales reps who rely heavily on email and might as well go home for the rest of the day if they can't get to their IMAP mailboxes while fsck does it's job for a few hours (on some of my systems using the older & slower 3ware driver that we talked about fsck would run for 4 or 5 hours, hence my choice of ext3). [1]
All my performance tests used ext2 as a base, but were also done with ext3 just to see what a difference it would make.
I should also mention it makes a BIG difference what block device your ext3 journal resides on.
My (informal) testing is that, compared to ext2, XFS *feels* slower, and JFS *feels* faster. I use neither, as NetBSD NFS clients don't seem to like 'em.
Depending on your tests, of course. I can understand how ext3 vs ext2 would make such a difference to dbench.In tests done recently Ext2fs gives more than 3 times the write I/O throughput of Ext3fs.
I didn't see much difference with bonnie.
[1] OK how's that for a long sentence :)
--
Phil Brutsche
[EMAIL PROTECTED]
--
To unsubscribe from this list go to the following URL and read the
instructions: http://lists.samba.org/mailman/listinfo/samba
