Re: Better cardinality estimate when dialing optimizer_features_enable back

  • From: Greg Rahn <greg@xxxxxxxxxxxxxxxxxx>
  • To: Neil Kodner <nkodner@xxxxxxxxx>
  • Date: Tue, 24 Nov 2009 11:28:40 -0800

So then it appears the predicate ("LETTER_BIN"."PRTY_ID"=102985) via
LETTER_BIN_PRTY_ID_I is row source with the issue.  In the first query
it looks to be off by 2 orders of magnitude, 1692 vs 14,  Thats
usually not so good in an NL plan.

Are the stats on the table and index both using 100% sample?
Is there a histogram on LETTER_BIN. PRTY_ID?
If you switch OFE to 9.2 just for the test query, do you get a
different cardinality estimate, or is it that OFE=9.2 just yields an
overall better plan for the full query?

On Tue, Nov 24, 2009 at 11:19 AM, Neil Kodner <nkodner@xxxxxxxxx> wrote:
> For what its worth, if I just explain plan on select * from letter_bin where
> prty_id=12345, I get an incorrect cardinality estimate.  For that reason, I
> have not tried isolating other fields.

Greg Rahn

Other related posts: