Re: What's the point of decreasing consistent gets by using IOTs?

"Jonathan Lewis" <jonathan_at_jlcomp.demon.co.uk> wrote in news:b7n7to$32b$1 $830fa78d_at_news.demon.co.uk:

>
> (a) Because a full tablescan starts with
> 2 (v9) or 4 (v8 - due to a bug) examinations
> of the segment header block before it starts
> scanning the data.
>
> (b) because the CPU cost of checking a
> condition on every row in a block could be
> much higher than the CPU cost of a binary
> chop through an index block to find a rowid
> and go to the right row in the table block.
>

Or, could the optimizer be reverting back to rules based if there are no statistics on the table? Received on Fri Apr 18 2003 - 15:44:55 CDT