RE: full-scan vs index for "small" tables
- From: "Laimutis Nedzinskas" <Laimutis.Nedzinskas@xxxxxxxxxxxxx>
- To: <oracle-l@xxxxxxxxxxxxx>
- Date: Wed, 28 Jun 2006 09:47:19 -0000
>On Behalf Of Nigel Thomas
>To be fair, the sentence continues (my bold, from 10gR1):
>"...a full table scan might be cheaper than an index range scan"
>or by implication it might not. Or they could be the same. Or England
could win the World Cup. Or whatever. Looks like they've got all the
angles covered there then.
And surely I am lost. What are the best practices for SQL programmer in
the modern post_Rule_Based_Optimizer Oracle then?
- Rely on unreliable CBO? But all we have already heard stories of even
Oracle fighting it's own CBO for *stable* data dictionary performance.
And CBO is inherently unstable at least in the beginning of
application's life cycle.
- Onlines? But they may or may not to cover all SQL. And outlines will
not cover literal SQL("legal" literal SQL), will they?
- Hint every single statement? That one seems to be the option, doesn't
it?
P.S. Just wandering: is there a hint to *disable* statistics at
statement or session level? I mean, let CBO use *defaults* making it
kind of rule based actually.
Thank you in advance,
Laimis N.
Fyrirvari/Disclaimer
http://www.landsbanki.is/disclaimer
Other related posts:
