Oracle FAQ | Your Portal to the Oracle Knowledge Grid |
Home -> Community -> Usenet -> c.d.o.server -> Re: To RAID or not to RAID (...or how to RAID)
Ed Stevens <spamdump_at_nospam.noway.nohow> wrote:
>
>RAID 5+0? Striped RAID-5? Never heard of these configurations. Can you point
>me to a description of this? RAID-5 *is* striped. The difference between 0 and
>5 is that 5 adds a parity check for drive recoverability. I'm trying to
>visualize what a doubly striped set would look like.
Essentially, visualise each RAID-5 set as a single device, not worrying about the internals any more than you worry about how a HDD distributes data between its platters. There's then just a RAID-0 stripe across N devices. It's not quite that easy, since you do have to be careful about stripe widths though, so you don't end up with suboptimal interactions.
The advantages of this is that you can have many disks in the logical volume, without the disadvantages of doing this in straight RAID-5. When you lose a disk in RAID-5, you have to read *all* the disks to reconstruct the data from the missing one, this is one reason why nobody likes very large RAID-5 sets.
The other is that they're not resilient against double failure. RAID 5+0 will survive a double failure if and only if the failures occur in different RAID 5 sets, which isn't as good as RAID 1+0, but isn't too bad. Just like every other use of RAID 5, its a compromise, better options exist but they use more disks, so cost more.
Hope that helps - Sorry, I've no references to hand and I'm sure you can use Google as well as I can. (I do however have a large RAID 5+0 configuration in the machine room here).
-- Andrew Mobbs - http://www.chiark.greenend.org.uk/~andrewm/Received on Fri Sep 13 2002 - 10:18:26 CDT