Re: why does CBO not use available indices

From: Noons <wizofoz2k_at_yahoo.com.au.nospam>
Date: Sun, 12 Oct 2003 01:58:15 +1000
Message-ID: <3f882873$0$28121$afc38c87@news.optusnet.com.au>

"Jonathan Lewis" <jonathan_at_jlcomp.demon.co.uk> wrote in message news:bm983d$djj$1$8300dec7_at_news.demon.co.uk...

> Oracle can decide to reverse the roles of the
> two tables in a hash join - in mid operation if
> the join is not optimal hash - the jargon for
> this is 'sides swapped' when it appears in the
> trace file.

Yikes! Never seen this one. Hope I never do...

> to be a carefully considered event. (In fact, the tests I've
> been running so far have been very small, and for a single
> statement I've been getting sensible results in less than
> a couple of seconds).

sweet! Now, THAT is what I call an improvement. Let's hope this 10g thing happens quick. Using 9ir2 at the moment, but next year I have a chance to push for 10g in the next project.

> statistics on a critical table, you get the effect you want. You
> upgrade from v8 to v9, and Oracle adds one to the cost of
> a table scan. Suddenly the costing on your 'pseudo-small'
> table behaves differently and every execution path on that
> table could change.

Agreed. It all has to be managed, until we get a more intelligent CBO. Let's hope 10g holds up well.

As a final note: some of the "Java heads" where I work are now using 9ir2 and JDev as well as Forms 9. To say they were surprised with the power of this combination to produce absolutely stunning applications is the understatement of the year. They still reckon J2EE is more scalable for some special cases. But for the vast majority of vanilla apps out there they agree Forms+9i or JDeveloper+9i is a much better solution. Ooooh YEAH!!!!

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

Received on Sat Oct 11 2003 - 10:58:15 CDT