SAME config

We have a brand new 2 CPU Sun V480, with a brand new StorEdge 3310, containing 12 72 G drives.

Thanks to this group, I found the Oracle S.A.M.E. paper, and I'd like to try it. Our main database has over 1000 tables, and similar # of stored procedures, but isn't really very large (< 10GB), and typically has fewer than 200 users, and <10,000 transactions per day.

The simplicity of SAME sounds very appealing.

I plan to use RAID 1 + 0 (mirror first, then stripe). So, I can have at most 6 mirrored pairs in my stripe. We may have to devote 2 disks for another, unrelated purpose, so I might only have 10 disks (5 mirrored pairs) with which to work.

So, 5 or 6 mirrored pairs, that are then striped.

The SAME paper suggests stripe width of 1 MB. My first thought is that that refers to each disk (pair) in the stripe. But that "sounds" larger than typical stripes. But then again, it seems like the SAME paper makes more sense that way, since on of its points is to minimize the seek time, relative to the transfer time. But then again, unless I am misunderstanding the Sun StorEdge 3310 documentation, the largest stripe width available on this particular hardware is 128 KB. A long way from 1 MB. Of course, even if the "1 MB" referred to the total of all disks participating in the stripe, I still couldn't make it 1 MB (limited to either 5 x 128, or 6 x 128).

Might someone please clarify what the precise intent of the SAME paper is?

For me, I guess either way I need to use the maximum (128 KB), unless there is some way to get a larger stripe chunk with the Sun StorEdge 3310.

Thank you.

-- 
David Grove

- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
"I think not," said Descartes, and disappeared.