I'm using estimate_percent=>100 and method_opt=>'FOR ALL COLUMNS SIZE 1' for these tables On Tue, Nov 17, 2009 at 3:27 PM, Greg Rahn <greg@xxxxxxxxxxxxxxxxxx> wrote: > First, are the cardinality estimates representative? > See "Using GATHER_PLAN_STATISTICS With DBMS_XPLAN.DISPLAY_CURSOR" > http://structureddata.org/2007/11/21/troubleshooting-bad-execution-plans/ > > If the cardinality estimates are off by a significant factor, then > validate your statistics. This may require a higher sample size than > what is currently used. > > There is no use in mucking with parameters/hints until the stats are > representative. > > On Tue, Nov 17, 2009 at 2:02 PM, Neil Kodner <nkodner@xxxxxxxxx> wrote: > > Once our environment was upgraded from 9.2.0.4 to 10.2.0.4, everything > > seemed ok except for UPDATE statements that contain correlated select > > statements. Across the board, we received poor performance compared to > > running them in 9.2. > > -- > Regards, > Greg Rahn > http://structureddata.org >