Re: Opinions on SAME?

It should really be MASE - mirror and stripe, not stripe and mirror. The implementation is a defensive one, i.e. a pretty good option when you can't be more fussy.

If you have only 6 discs, it is almost certainly better to have 3 pairs striped with redo logs smeared in, rather than 2 pairs striped and 2 separate spindles for redo.

If you have 100 discs, it is far more sensible to keep 2, or 4 discs out for redo, then mirror and stripe the rest. But in this case you wouldn't want to stripe across 48 pairs, you would be far better off creating (say) 6 sets of 8 pairs, or maybe 3 sets of 16 pairs.

SAME is really aimed at the smaller systems which can fit comfortably on a small number of spindles.

--
Jonathan Lewis
Yet another Oracle-related web site:  http://www.jlcomp.demon.co.uk

Practical Oracle 8i:  Building Efficient Databases
Publishers:  Addison-Wesley

Reviews at: http://www.jlcomp.demon.co.uk/book_rev.html



Steve Cole wrote in message <3ADD8B97.72336EEB_at_fmgtech.com>...


>If anyone has any firsthand experience implementing this or has a strong

>opinion one way or the other please share.

>

>I agree in principle with most of it - the more physical disk heads you

>can keep working for you the better off you are, and RAID 0+1 is faster

>than RAID5.

>

>I have to question putting the redo logs into that mix though. Redo is

>sequential writes, the other components are random access. I have to

>think that keeping the redo on a separate hardware RAID 1 by itself

>would be faster than putting it on the RAID 0+1 with everything else.

>

>Thoughts?