| Oracle FAQ | Your Portal to the Oracle Knowledge Grid | |
 |
 | |||
| Oracle FAQ | Your Portal to the Oracle Knowledge Grid | |
|
| |||
Home -> Community -> Mailing Lists -> Oracle-L -> Re: 10046 event is useless in 9.2.0.4 version for response time measuring !!!
If you can extract the performance problem from production and replicate it on a non-production system, is the 6x timing issue
really an issue? If the 6x stays constant, then you can still compare apples to apples. You can hypothesize the results by factoring
in the 6x into your tuning forecast.
Query 1 in production takes 10 minutes
Query 1 in staging w/10046 takes 60 minutes
Tuned Query 1 in staging w/10046 takes 6 minutes
Tuned Query 1 in production *should* take 1 minute
With a 6x impact, I certainly would not want to be running 10046 in production.
<humor>
Then again, it is a great opportunity to impress the users. Turn on 10046 in production, watch performance degrade by 6x. Turn off 10046 and watch it improve. Users are happy and totally impressed with your handling of the problems. Of course, to make yourself even more valuable, you could randomly turn on 10046 for a session, wait for the phone call, take a 'smoke break', turn off 10046 and then call the user back in an 'overworked' voice.
You can also impress management. Base line the 6x query in staging, tune it a little, then show management the 12x - 120x improvement in production.
</humor>
Cary Millsap wrote:
> Jurijs,
>
> Even if something takes 6x longer to run the one time you trace it, it's
> still better to have the detailed 10046 data than risk being tricked by =
> the
> aggregation problems introduced by using V$ data.
-- Archives are at http://www.freelists.org/archives/oracle-l/ FAQ is at http://www.freelists.org/help/fom-serve/cache/1.html -----------------------------------------------------------------Received on Mon Aug 02 2004 - 15:26:34 CDT
 |
 |