Re: Oracle benchmark

On Fri, 03 Mar 2006 11:38:02 +0100, Michele Campagni wrote:

>
> I did many tests but never be able to stop I/O access. But if you say SGA
> never contains entire DB I belive it.

There are many, manyu reasons for the traditional Oracle database instance to be accessing the disk during a query. Not all relate to executing a query.

Oracle, like all vendors, spends literally thousands to millions of dollars tuning TCP benchmarks to get minute performance gains. This is done by people who live and breathe Oracle tuning - and attempting to blindly put the entire SGA into memory is NOT one of the things they typically will do. (They will make judicious user of the SGA, and the various pools.)

If the goal is truly to find the relationship between CPU and performance, you need to study the Parallelism apabilities of Oracle very carefully. From what I can tell, Oracle uses several advanced algorithms to parallelize operations.

However, I agree with Frank that a first cut using TimesTen is a great start down an alternate path.

-- 
Hans Forbrich                           
Canada-wide Oracle training and consulting
mailto: Fuzzy.GreyBeard_at_gmail.com   
*** Top posting [replies] guarantees I won't respond. ***