On 04/01/99 22:14:49 you wrote:

On Thu, 1 Apr 1999, Val Bercovici (NetApp) wrote:

...

5 data disks or 5 disks including parity (meaning 4 data disks)? Either way, that's well below our sweet spot of 14 9GB disks per raid group.

Wow, I'm discovering my psychic abilities. 14 is exactly the raid group I chose on our filers. Can you share some data on how NAC arrived at this number. In our case it just happened that we wanted to have two volumes as separate as we could make them and we happened to have 4 shelves in each filer.

I think it's derived empirically, although you can make some rough calculations yourself and see when you have a typical "mix" of NFS operations (reads, writers, lookups, getattrs, etc.) it's hard to get more than 100-200 NFS ops per disk, with the filer being in the upper end of that range. Of course as disk drives improve this should as well. Unless you want to set off a very small filesystem that's only going to get limited access, you probably want more disks than that.

In other terms of raid group size, I think 14 became an ideal spot because it was a good compromise between low overhead (the more data disks per parity disk you have, the less disk you waste), acceptable raid reconstruction times in case of disk failure (the fewer disks you have, the quicker the reconstruction), and ease of use (it's conventient to think of disks in groups of 7 and 14).

Bruce