Never RAID 5 for a database. When I say "never," I can give you edge-case scenarios, but you're basically taking a 4x overhead on all writes. Now, RAID-0 is a bad choice as well, since JBOD has no replication, but it sounds like you might have that end under control.
-------- Original message --------
From: Rene Romero Benavides <rene.romero.b@gmail.com>
Date: 08/27/2013 5:35 PM (GMT-06:00)
To: pgsql-admin@postgresql.org
Subject: [ADMIN] 3 disks configured RAID 0 over 10 disks configured in RAID 5 (self replicating SAN)
Thanks for your attention.
What would you choose for a postgresql installation: 3 disks configured with RAID 0 (in a self replicating SAN) over 10 disks configured with RAID 5 (also in a self replicating SAN) , we have space constraints that prohibit us from choosing RAID 1+0.
I've been persuaded to choose RAID 5, because writes and parity computation will be spread over 10 disks compensating write overhead providing a better level of data security.
Do you think it was a good decision? Any comment will be appreciated. Have a good day.
--
El genio es 1% inspiración y 99% transpiración.
Thomas Alva Edison
http://pglearn.blogspot.mx/
