On Wed, Jun 11, 2003 at 03:06:37PM -0600, Chris Jones wrote:
> Manuel Bouyer wrote:
> >What is the output of raidctl -a, and the disklabel of your raid ?
> 
> I assume you mean raidctl -s, not -a.

Yes, sorry for the typo

> Chris Jones               chris@cjones.org                www.cjones.org

> Components:
>            /dev/wd2e: optimal
>            /dev/wd1e: optimal
> No spares.
> Component label for /dev/wd2e:
>    Row: 0, Column: 0, Num Rows: 1, Num Columns: 2
>    Version: 2, Serial Number: 2002052458, Mod Counter: 345
>    Clean: No, Status: 0
>    sectPerSU: 32, SUsPerPU: 1, SUsPerRU: 1

This means 16k per SU, so I/O to the disk won't be larger than 16k

> # /dev/rraid0d:
> type: RAID
> disk: raid
> label: default label
> flags:
> bytes/sector: 512
> sectors/track: 32
> tracks/cylinder: 8
> sectors/cylinder: 256

This is probably suboptimal: with sectors/cylinder a multiple of sectPerSU,
the metadata may all end on the same disk (I mean, all end on a odd or even
number of sectPerSU).
I would recommend rebuilding the raid with sectPerSU set to 64 instead of 32
(to have larger I/O to disk) and rebuild the filesystem with a non-power-of-2
as sectors/track and/or tracks/cylinder.
I usually use: 64 sectPerSU, 63 sectors/track, 16 tracks/cylinder.
I also bump cpg at newfs time, but I'm not sure it makes a difference for
performances (it just wast a little less space).

-- 
Manuel Bouyer <bouyer@antioche.eu.org>
     NetBSD: 24 ans d'experience feront toujours la difference
--