RE: CBO - hash join vs nested loops

From: oracle-l-bounce_at_freelists.org
[mailto:oracle-l-bounce_at_freelists.org] On Behalf Of Radoulov, Dimitre

>>1.With apparently correct row source estimate, the CBO chooses not
optimal
access path.

>I've already read chapter 5 for the clustering factor, congratulations
for
the great work!

I am interested to here the conclusion: was Oracle CBO correct or not? I see the following options:

1. CBO was just wrong
2. Oracle was wrong because of inaccurate statistics
3. Oracle was correct.
4. One interesting option: currently Oracle is correct but in the future(when more data comes into the system) it can possibly fail. In theory this case should be handled by regular statistics gathering but in practice it depends on the answers to 1-3.

--
http://www.freelists.org/webpage/oracle-l