On Tue, Jul 22, 2008 at 02:42:28PM -0700, Sumit Gupta wrote:
> Hi Bob
> 
> For a storage subsystem 'COMSTAR in itself' is only a component. Other 
> components are HBA, PCI, CPU, backend storage etc. Each one of those 
> influence the performance.  But so far here is what we have found. Our 
> reference platform is x4500 (Thumper) and we use zfs raidz zvols as our 
> backend. Also the protocol for these tests is FC using Qlogic 4G HBAs.
> I have to also note here that most of this is done as a part of our 
> performance regression testing, the purpose of which is to make sure 
> that any bugfixes we make don't regress the performance. There is no 
> formal performance data yet.

I'd be interested in any recommendations you might have on raidz
configurations in the x4500.  I'm looking at deploying COMSTAR+ZFS from
x4500s via our current SAN infrastructure in the very near future
(currently trialling with a pair acquired on TryNBuy).  Currently I'm
using SXCE b91 and this RAID setup (note that our boot disks are on c6
because we installed from a USB DVD drive):

for i in 1 2 3 5 6 7; do zpool create poolt$i raidz c0t${i}d0 c1t${i}d0 
c5t${i}d0 c6t${i}d0 c7t${i}d0 c8t${i}d0; done
for i in 1 2 3 5 6 7; do zpool add poolt${i} spare c7t0d0 c8t0d0 c7t4d0 c8t4d0; 
done
zpool create pool_r10 mirror c0t0d0 c1t0d0 mirror c0t4d0 c5t4d0 mirror c1t4d0 
c5t0d0

Performance on the raidz sets looks to be within about 9% of our EMC
storage for most workloads, and I'd love to close that gap a little
if possible.

Ceri
-- 
That must be wonderful!  I don't understand it at all.
                                                  -- Moliere

Attachment: pgpTFIvoY3AIU.pgp
Description: PGP signature

_______________________________________________
storage-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/storage-discuss

Reply via email to