Question re System Statistics

Greetings,

I would be interested in folks thoughts about how frequently system statistics should be gathered and on what criteria should be used to make this determination. My thoughts are that this is dependant on the environment but I'm not sure to what extent this is so. My expectation is that the system metrics as reflected would remain fairly consistent over time but I may be way off base here. This is based on my assumption that the hardware more than the code is the determining factor here. Once the code knows how the hardware is behaving it will take that into consideration in building the execution plan (I'm not sure I'm stating this clearly). So the way I am understanding things one really only needs to gather system statistics once initially and then leave them alone. But then that leaves the question, and this is where I get a bit confused, if I were to run a very well tuned efficient piece of code and gather system statistics while it is running and then run a horribly inefficient piece of code and gather system statistics while it is running would I see great differences in the system statistics (I haven't tried to test this because I only write well-tuned highly efficient code <g>). I think I am sort of asking if really horrible code can give a bad view of system statistics and is it really necessary to regather system statistics on a regular basis. I have never really been able to find an explanation for this and am looking for some guidelines.

Thanks.

Bill Wagman
Univ. of California at Davis
IET Campus Data Center
wjwagman_at_ucdavis.edu
(530) 754-6208

--
http://www.freelists.org/webpage/oracle-l