On Mon, Aug 3, 2009 at 4:02 PM, Rodrigo E. De León Plicet<rdeleonp@xxxxxxxxx> wrote: > On Mon, Aug 3, 2009 at 11:15 AM, Brian Modra<brian@xxxxxxxxxxxxx> wrote: >> (...) and 6x300GB SAS RAID 10 for the database... but some experts have said >> RAID 5 is fine. I'm inlined to think RAID 10, but I'm not an expert. > > These guys: > > http://www.baarf.com/ > > ... kinda dislike RAID5 with a passion; dunno if it's related to stuff like: And understandably so. RAID5 was invented back when hard drives were measured in megabytes, and not necessarily hundreds of them either. Nowadays, you've got two very different uses for arrays of drives. One is to provide a LOT of storage for a reasonable price, and the other is to provide maximum throughput by aggregating large numbers of drives together. Where I work we have both. We have media servers that are running 8 2TB drives in a RAID-6 array to provide 12 TBs of redundant storage. We also have primary db servers running 12 140G SAS drives in a RAID-10 for fast access storage, providing only 850G or so of storage. This is for a db that uses just under 100G of drive space. It's no where near the maximum of that array, and long before we run out of space we'll be adding more drives /controllers to keep up with the performance needs of the db. RAID-5/6 would make no sense there whatsoever. If you HAVE to go with a striping type solution, RAID-6 is generally better than RAID-5. It's like RAID-5 with the spare drive already built in and generated, and it behaves much better with a failed drive than RAID-5, and is much less likely to suffer from data loss, requiring three drives to fail in order to lose coherency. RAID-5 is the worst of all compromises. -- Sent via pgsql-admin mailing list (pgsql-admin@xxxxxxxxxxxxxx) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-admin
