Those setting partially cancel each other out: the db_file_multiblock_read_count value of 32 causes the optimizer to assign a lower cost to full table scans and subsequently to hash joins. The setting of optimizer_index_cost_adj to 30 causes the optimizer to assign a lower cost to index accesses and to NL joins.

Given the setting of optimizer_index_cost_adj I would guess that your FK_CF02CON_TSA index has a high clustering factor. As I said in my previous post, try deleting the statistics for that index and see what happens.

How come that statement uses some bind variables, but then one literal? If it were all bind variables you could lock in the good plan with a stored outline - for example by deleteing the statistics on both tables, which will cause the RBO to parse the sql

At 05:49 PM 9/10/2003 -0800, you wrote:
Tks Wolfang

I have read the paper.

db_file_multiblock_read_count        integer     32
hash_area_size                       integer     4194304
sort_area_retained_size              integer     1048576
sort_area_size                       integer     2097152
hash_multiblock_io_count             integer     0
optimizer_index_caching              integer     90
optimizer_index_cost_adj             integer     30

Ramon E. Estevez
[EMAIL PROTECTED]
809-535-8994

Wolfgang Breitling Centrex Consulting Corporation http://www.centrexcc.com

--
Please see the official ORACLE-L FAQ: http://www.orafaq.net
--
Author: Wolfgang Breitling
 INET: [EMAIL PROTECTED]

Fat City Network Services    -- 858-538-5051 http://www.fatcity.com
San Diego, California        -- Mailing list and web hosting services
---------------------------------------------------------------------
To REMOVE yourself from this mailing list, send an E-Mail message
to: [EMAIL PROTECTED] (note EXACT spelling of 'ListGuru') and in
the message BODY, include a line containing: UNSUB ORACLE-L
(or the name of mailing list you want to be removed from).  You may
also send the HELP command for other information (like subscribing).

Reply via email to