Re: How to tune for "too many" users ?

Hi, Stephan, thanX for your input. comments inside

"Stephan Bressler" <stephan.bressler_at_siemens.com> schrieb im Newsbeitrag news:b8o0v8$qk1$1_at_news.mch.sbs.de...
> Hi,
> here some ideas:

> - check memory (avoid heavy memory management) and cpu usage on OS level.
> - check for disk io

Memory usage and OS-IO going up only at very heavy duty times while processing the shipping of products, daily about 1-2 hours, which makes redologs rotating withing minutes (with respect to hours at others times). This bottleneck is well
expected and so far cannot be avoided since many data are involved due to the data model. But this has been the same situation since I came in and before that. And then, there were no jams like I can see today.

Seems like something changed just over the last 4-6 weeks, but the last parameter change was 4 months ago.
> - optimizer mode = choose.

parameter is set to choose, and we use optimizer hints.

> Did you generate statistics on tables / indexes?
Yes, last_analyzed just 3 weeks ago.

> - did you use CBO with 8.0, too?

Yes, we did.

> - check for most expensive statements in terms of diskreads and buffergets
> (v$sql). Tune them!

I already was searching in v$sql and so on, and this is certainly one important point.
I find many expensive statements (thatīs why I said dredging the ocean), but this is not
much different from the normal situation (the inhouse app is our own developement and we are
almost constantly working to make it better). So I thoung maybe there is another source of problems,
maybe the SGA, but I am not sure ... Received on Wed Apr 30 2003 - 04:27:55 CDT