Measuring database response time

Subject: Measuring database response time

Platform: Ora 8.0.5 SE, NT 4 sp6

While I am making reasonable process on identifying performance issues with inefficient SQL, need for judicious use of indexes, and convincing the apps programmers to use bind variables, I have a growing suspicion that most of our response time issues are outside of the database. My current problem is a browser-based app that users are reporting a response time of 5 to 15 minutes with frequent time-outs. What I'd like to find out is how much of that response time is due to database or SQL issues, and how much is completely outside of the database's control.

I really like the summary that Oraperf provides, and have been feeding it a lot of estat reports. Those of you familiar with Oraperf know that it has a section called "Response Time" with breakouts of CPU and Wait times - per execute, per user call, and per transaction. At first I focused on the "per transaction" numbers, but after comparing the Oraperf report with the original estat report, it appears that
"per transaction" would be more accurately labeled as "per sampling
period."

So . .

Can someone explain what the Oraperf reports mean by "per execute",
"per user call" and "per transaction"?

And . .

Is there some number I can derive from the estat report that I can show the programmers and managers and say with confidence "this clearly shows that Oracle is turning a request in x seconds, so if you have a response time of 20x, you need to be looking elsewhere."

--
Ed Stevens
(Opinions expressed do not necessarily represent those of my employer.)