Re: CBO - hash join vs nested loops

  • From: "Radoulov, Dimitre" <cichomitiko@xxxxxxxxx>
  • To: <lnd@xxxxxxx>
  • Date: Mon, 5 Dec 2005 15:21:51 +0100

- Should those queries be written such that CBO is forced to take what the developer thinks is an optimal plan? Interestingly, here the old good NL-loop turns out to be a better plan than hashing, scanning, etc.

- Is the first thing to do with the Oracle instance that to tune OCI? Should high OIC values (the "correct 99.9) be tested immediately? (This is actually what I do - set all OLTP or general purpose instances to favour indexes.)

I think the correct approach is to tune immediately the OIC, but for that I have to verify the behaviour of all the problematic statements with the new settings.
I repeat, I have an issue that looks very similar on more than one 9ir2 instance.




Regards,
Dimitre Radoulov

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


Other related posts: