Re: "'Resultcache" effect in 10.2.0.4 ??
Date: Mon, 27 Jul 2009 23:17:26 +0200
"jlchiappa" <jlchiappa_at_gmail.com> wrote in message news:68e2e528-9627-4e13-bf85-372283ee2fc9_at_a37g2000prf.googlegroups.com... Mark, I disagree : as I said, if the exact same SQL is run with the exact SAME values 2x (thus NEGATING the possibility of bind variable peeking, the value IS the same for both executions), the first execution is slow and the second is fast... Until now, the only hypothesis explaining the situation really is some kind of cache outside the db (in storage, in file system, whatever), when I get a 10046 trace I will to comprove it, if contrasting 2 executions with the same plans and the same I/Os the 2nd is faster the time-elapsed for the I/Os in the 2nd will be smaller, thus comproving the acting of some "force" outside the db (hardware caching, file system, any). Only to post, my plan of action will be : trace the SQL´s run for the 1st time, clearing all dbcaches , invalidate SQL, etc, after this, and then trace a 2nd execution...
Isn't this normal Oracle behaviour? The blocks are retrieved from disk and cached in the SGA on the first execution and retrieved from memory on the second execution?
Matthias Received on Mon Jul 27 2009 - 16:17:26 CDT