Hardware

I hope I am posting this in the right place.

I would like to gain some further insight into the notion that a Raid 10 is always faster than a Raid 5 or 6 for database writes and reads.

And should be used for database deployments even if 75% + of database activity is reads.

The array I am looking at is a 14 disk array on HP MSA 1000 with dual controllers, the disks are 300gb 10,000R RPM utlra320 SCSI's.

This unit currently has six luns on it. All striped across the 14 installed disks in raid 6 configuration.

In the array 3 luns are for our production eBusiness suite Apps and DB tiers and 3 luns are for our eBusiness Suite test/development apps and db tiers.

I am considering whether or not converting to a raid 10 would gain enough performance improvement to justify the loss in disk space.

The question I have is how can a Raid 10 "two sides of a mirror with 7 disks striped on each side of the mirror" be faster than a Raid 6 which is "14 discs in the stripe with 2 parity blocks".

Even if you drop 2 disks out of the Raid 6 stripe for parity you are still striping across 12 disks once verses striping across 7 disks twice.

I recently saw an article that made the case that once you pass 8 disks in an array of disks that Raid 5 and 6 will out perform Raid 10 as long as the control unit for the array is a quality unit like an HP MSA unit.

Thanks in advance for any and all help with this question.

Russell.