Re: New Connection for Each Query
Date: Wed, 11 Jun 2008 08:38:35 -0700 (PDT)
Hi Kyle & Mark,
Thanks for the ideas. Unfortunately this is a legacy app (on Oracle 9i) with no "switch" to turn on connection pooling. Kyle - I'll try your suggestion of running nothing but logons at the current rate (only 2 per second but the CPUs are slow anyway) to see what load I get. Can't do this easily just now due to constant use of the system but hopefully get time to try it later. Thank you!
- Original Message ---- From: "kyle Hailey" <kylelf_at_xxxxxxxxx> * To: mark.powell_at_xxxxxxx * Date: Mon, 9 Jun 2008 17:22:46 -0700 Probably the best method is just to measure it yourself if you can. Write a script that connects/disconnects at a high rate and measure the CPU effect on the machine when nothing else is running. Run the script at the logon rate of your current applciaiton. You can get the logon rate from statspack.
If you are on 10g you can look at the time model for connection time , ie
"connection management call elapsed time" in view v$sys_time_model.
Look at this with AWR report or Statspack to get the delta and see
how much time you spend connecting much/most of which can be CPU, but
I've had smaller than expected values from this statistic before
(1/10th of what I expected) so I just set up a script to logon/logoff
when nothing else was going on the machine and measured how much CPU
was being used. It's quite expensive. In my case , if I recall
correctly, 4 connections a second was a full CPU the machine.
http://www.freelists.org/webpage/oracle-l Received on Wed Jun 11 2008 - 10:38:35 CDT