Comments on: Side effect of cursor sharing http://laurentschneider.com/wordpress/2008/04/side-effect-of-cursor-sharing.html Oracle Certified Master Sun, 07 Sep 2008 13:29:07 +0000 http://wordpress.org/?v=2.0.5 by: Gary http://laurentschneider.com/wordpress/2008/04/side-effect-of-cursor-sharing.html#comment-6146 Fri, 04 Apr 2008 01:07:02 +0000 http://laurentschneider.com/wordpress/2008/04/side-effect-of-cursor-sharing.html#comment-6146 Interesting quote from the Oracle optimizer people http://optimizermagic.blogspot.com/2007/12/why-are-there-more-cursors-in-11g-for.html "We hope that in the future, this feature [11g's adaptive cursor sharing] will persuade people to set cursor_sharing to force." I wonder if there is, or will be an event, to record the bind values used when a query is optimized. Interesting quote from the Oracle optimizer people
http://optimizermagic.blogspot.com/2007/12/why-are-there-more-cursors-in-11g-for.html

“We hope that in the future, this feature [11g’s adaptive cursor sharing] will persuade people to set cursor_sharing to force.”

I wonder if there is, or will be an event, to record the bind values used when a query is optimized.

]]> by: Laurent Schneider http://laurentschneider.com/wordpress/2008/04/side-effect-of-cursor-sharing.html#comment-6144 Thu, 03 Apr 2008 11:36:42 +0000 http://laurentschneider.com/wordpress/2008/04/side-effect-of-cursor-sharing.html#comment-6144 Thanks for the comment. No, it was not trace files with bind. About bind peeking, there is another value SIMILAR which partly solves the problem. Still I set the parameter back to EXACT on that db. Thanks for the comment.

No, it was not trace files with bind.

About bind peeking, there is another value SIMILAR which partly solves the problem. Still I set the parameter back to EXACT on that db.

]]> by: Yas http://laurentschneider.com/wordpress/2008/04/side-effect-of-cursor-sharing.html#comment-6143 Thu, 03 Apr 2008 11:07:54 +0000 http://laurentschneider.com/wordpress/2008/04/side-effect-of-cursor-sharing.html#comment-6143 If you have an sql trace with binds (10046 with level 12) you have the chance to find out the values. It may be hard if there are hundreds of different values, but you should not be using literals for such a query anyway, so there is no difference here for sqls with hundreds of different bind values. Laurent, another side effect (and a more important one) is bind-peeking. Before you had optimal execution plans(thanks to the literals), now you have binds, so bind-peeking. If there is a histogram on the column you may get suboptimal plans for some bind values and optimal plans for others. If you have an sql trace with binds (10046 with level 12) you have the chance to find out the values. It may be hard if there are hundreds of different values, but you should not be using literals for such a query anyway, so there is no difference here for sqls with hundreds of different bind values.

Laurent, another side effect (and a more important one) is bind-peeking. Before you had optimal execution plans(thanks to the literals), now you have binds, so bind-peeking. If there is a histogram on the column you may get suboptimal plans for some bind values and optimal plans for others.

]]>