>I think that if your miss of sql in the shared pool is 10% that most >statements that are not frequently executed will quickly be aged out >and re-parsed during subsequent execution. How about great guideline - parse ones execute many times. In this context, the systems with good design which use bind variables are more instable then others ;) >How is this any different, if the bind variables used at parse (with >cursor_sharing=SIMILAR) differ greatly from one execution to another, >in terms of distribution? And this is just begining of the list: - PGA_AGGREGATE_TARGET - bind variables peeking - regular statistic gathering ... ... ... More features we have (use), more instable our systems ;) But there are some methods to get rid of it, or at least minimize impact I suppose. Jurijs On 07.10.2004 03:05:28 oracle-l-bounce wrote: >Jurijs, > >I think that if your miss of sql in the shared pool is 10% that most >statements that are not frequently executed will quickly be aged out >and re-parsed during subsequent execution. > >If then follows that this is only a concern if your application SQL >code is nearly perfectly written and statements are not >invalidated/aged out. > >Right. > > >How is this any different, if the bind variables used at parse (with >cursor_sharing=SIMILAR) differ greatly from one execution to another, >in terms of distribution? > >Paul > > -- //www.freelists.org/webpage/oracle-l