Re: SQL Tuning Regarding System CPU Stats

On Fri, 18 May 2007 07:36:13 -0700, andrew.markiewicz wrote:

> I frequent Tom Kyte's site and have both his books, studied and
> restudied Wolfgang Brietling's paper on the 10053 trace such that it is
> tattered, highlighted and stained with coffee and tears, own and read
> Jonathon Lewis's "Cost Based Oracle Fundamentals", and peruse OTN,
> Metalink, groups such as this and other forums. I am familiar with HOW
> to tune, but was interested in how others approach setting up the
> environment for tuning with using system statistics.

Andrew, you are far too aggressive with the system statistics, that is why I thought that you were a beginner. System statistics is just a benchmark telling the CBO how many I/O requests can your system do and how fast is your CPU. Problem with the aggressive statistics computation is that you are introducing instability into the system, as the new stats have potential to change plans. As for batch processing, it is always desirable to do it on a separate machine.

-- 
http://www.mladen-gogala.com