Re: SAME config

I certainly thank you for your explanation, Hans. I appreciate the point that it isn't really necessary to agonize over placement of data, indexes, etc., and that we can obtain most of the benefit of doing that with a simple "one-size-fits-all" scheme which is known to be suboptimal, yet is "close" to optimal, and very robust.

I didn't ask my question clearly. I was mainly wondering about the meaning of "stripe width" in the SAME paper. My first impression is that it means the size of the stripe on each individual disk. (Thus, with 6 disks, the total stripe size would be 6 MB). It just seemed a little large. But, then that is consistent with causing the seek time to be small with respect to the transfer time.

Thank you, again.

DG

"Hans Forbrich" <forbrich_at_yahoo.net> wrote in message news:eh2Zc.37211$A8.16572_at_edtnps89...
> David E. Grove wrote:
>
> >
> > Might someone please clarify what the precise intent of the SAME paper
is?
> >
>
> An acknowledgement that the 'traditional' administration of Oracle
databases
> by separating segment types and further separating disparate table 'types'
> to optimize disk-related performance by managing disk-head movement can
now
> be relegated to myth status.
>
> IOW - there is little/no need to put indexes, look-up tables, heavily used
> transaction tables and so on into different tablespaces, and separate
those
> tablespaces to different disks or disk controllers, *for the purpose of
> optimizing performance*.
>
> (There are other, valid reasons for such separation, but performance is
not
> one of them.)
>
> /Hans
Received on Tue Aug 31 2004 - 12:28:13 CDT