Re: optimizer_ ???

I attended Wolfgang's presentation on oica and oic at hotsos, and towards the end of the presentation, he said using system statistics would provide better results than adjusting oica and oic. This was consistent with what I found last summer although I didn't realize it at the time. (Unconsciously competent? hope so - it's better than the alternative)

I had begun testing some of our key queries with changes in the optimizer_index_cost_adj and noticed that some queries did improve slightly. Then I got sidetracked and ended up implementing 2 sets of system statistics; one for working hours and one for batch processing.  When I finally picked back up on testing the oica changes, I didn't notice the same levels of changes in the queries. I figured I had misremembered my results, until I heard Wolfgang's presentation.

Of course, if I've misremembered the presentation, someone please let me kn= ow.

Robyn

On 4/22/05, Post, Ethan <Ethan.Post_at_ps.net> wrote:
> Based on W.Breitling's observations about oic and oica your settings
> might be just fine (I was at Hotsos but did not hear this talk, but that
> is one thing I recall a friend saying). The other settings look default
> also but would have to check on dynamic_sampling.=3D20
>

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