Ext3 and 3ware RAID5

[Joshua Baker-LePain]

On Sat, 15 Apr 2006 at 12:18pm, Alex Bligh wrote

>
>
> --On 14 April 2006 18:31 -0400 Joshua Baker-LePain <jlb17 at duke.edu> wrote:
>
>> Any ideas as to what leads to ext3's performance hit?  I've tested *lots*
>> of configurations.  It's not the md layer -- 1 card setups see the same
>> performance hit.
>
> No idea, but I suffer the same problem with a 9550SX-4 with SATA3
> drives. I don't think ext3 is the problem as dd gives much the same
> behaviour (you might want to try it). My reluctant conclusion is
> that the 9550SX is just dog slow, which is why the drives are currently
> sitting idle. I would love someone from 3ware to disprove this.

When ext2 is almost 2X faster than ext3 at writing, it points pretty 
firmly at something in the journaling code (IMO).  And a journaling FS can 
do decently with this hardware, as XFS also gets ~300MB/s writing.

-- 
Joshua Baker-LePain
Department of Biomedical Engineering
Duke University