Re: cpu time and query column in tkprof output

Hi,

> If both do the same, shouldn't at least bytes sent be the same?

They don't necessary do the same. Correlated EXISTS and a join return different results if the join key (in this case faxutil.t_kfs_tabstatuses.tabno , faxutil.t_kfs_tabstatuses.pool_status, faxutil.t_kfs_tabstatuses.job_status) is not unique.

In general I would try to avoid those execution plans:

FILTER (in select 1)

   something

   FULL something

NESTED LOOPS (in select 2)

    something

    FULL something

Those executions are not very scalable in terms of the cardinality of the user row sources.

My personal preference is, if there is at least one hash join in the execution plan, I switch all joins to [semi/anti] hash.

Regards

Jaromir D.B. Nemec

http://www.db-nemec.com

• Original Message ----- From: "Wolfgang Breitling" <breitliw_at_centrexcc.com> To: <yasbs_at_kocbank.com.tr> Cc: <oracle-l_at_freelists.org> Sent: Wednesday, February 02, 2005 3:34 PM Subject: Re: cpu time and query column in tkprof output

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