Re: Separating data, index objects

>So even in a *single* user system, separating indexes and tables provides
>*no* performance benefit. In theory or not.

OK, I don't recall the older threads response to my question.

Given: Newly formatted large disk. Batch job, single user, full table scan (ie, end of year report). One table in the tablespace on the device. One index, not used here. Low pctfree. Fair-sized multiblock read count. Continuously increasing primary key. OLTP system, in general.

Are you saying there will be no performance difference based on these two possibilities:

1. The table was loaded by an import, so has all blocks contiguous (well, in a few chunks from what I've seen), then the index was created in another tablespace.
2. The table was loaded by a process that added each row and its index within discrete transactions, and the index is in the same tablespace.

It would seem to me #1 would have a higher desired information density, or at least the multiblock read would have a higher chance of not wasting time on an index block.

jg

--
@home.com is bogus.
CalPers is buying land in China and China is buying Unocal and Chinese
business people want to buy WalMart.  The world is truly going insane.