RE: Extended trace analysis and perf. problem.

Sorry - the number of events per run will be different - it was just a sampling of a run - for about 10-15 minutes per run. A 10-15 min run generates about a 30M logfile so I was trying to keep things manageable. I was basing my wait assumptions on the averages from each run not on the number of events.  

• Brian

Christian Antognini <Christian.Antognini_at_trivadis.com> wrote: Brian

>Is it possible to change anything else? :-)

Yes, database version.

>All of the wait time averages for the new setup=20
>are right in line with current production.

Response time =3D service time + wait time

What about CPU?

>If this was database related I'd expect to see=20
>some wait-time averages way off the chart somewhere
>- wouldn't you? Am I missing something if this is=20
>database related - shouldn't it show in the extended
>traces if it was?

If the loads are about 1/2 fast, the response time on the new system = should be twice as the old system. Therefore either you use more CPU or = you wait more time! (or both of course)

>RAC 'C' is the new cluster - all the others are=20
>current production runs.

Mhmm... without analyzing the data in detail, if the forth column is a = run, to me it seams that there are huge differences!

HTH
Chris

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


__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com 

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