Re: INDEX and DATA Tablespaces.....

"Daniel Morgan" <damorgan_at_x.washington.edu> wrote in message news:1063812602.124378_at_yasure...

> I understand what you are saying but in a multiuser environment which
> came first the index block read or the table block read is reasonably
> irrelevant: Both are going on simultaneously. Then throw in some inserts
> and updates and the party starts to get interesting.

Exactly. The separation may be relevant in a single user dedicated situation, but even there I doubt it.

Once we throw in the randomizing effects of multiuser TOGETHER with the in-built disk cache (which you get like it or not), plus any controller cache, plus OS cache, plus Oracle cache, to claim that ANYTHING happens in ANY sequence is IMHO wishful thinking.

> there. I think the best real world solution is SAME. Then it all becomes
> irrelevant.

Same here. (no pun intended!)

It is a relatively easy to implement solution that finds a middle ground and evens out ONE of the many variables: the I/O response. Letting the DBA then concentrate on all sorts of other links along the chain with relative assurance I/O speed is not suddenly gonna drop off because of some unforeseen circumstance.

Of course, if one is after the last millisecond of I/O capacity and speed, then SAME is not the way to go. It's all relative.

-- 
Cheers
Nuno Souto
wizofoz2k_at_yahoo.com.au.nospam