Re: Oracle Benchmark Results for Different Hardware Configurations?

Serge Rielau wrote:
> Bob,
>
> Do you look at "miles/gallon" when you buy a car? Or perhaps you look at
> the safety rating. How is that different?
> You use controlled tests to make _relative_ comparisons and get an
> _indicator_ of expected behavior.
>
> Naturally the closer the test is to your real situation the more
> accurate it will be. But the closer it is to your configuration the more
> likely it is that you had to invest either your time or your money or
> both to get it done.
>
> Cheers
> Serge

Impressive analogy. The difference though is that those safety tests are conducted by an independent organization ... not by Ford, Toyota, etc. and there is no option for them to alter their vehicles in order to gain an advantage over a competitor.

In the case of these trumped up benchmarks ... games are played by the vendors that do not reflect anything remotely close to the real world. If you tell me RDBMS alpha running on hardware platform beta with operating system gamma gave a specific number that means nothing if I don't have the ability to drill down and figure out that you used specific object types and specific tuning tricks to enhance the number. And not one vendor ever advertises how they managed to pull the rabbit out of the hat.

I see the entire exercise as basically dishonest. I'd rather take my application to the lab at hardware vendors beta and delta and find out what is real. But I guess that wouldn't make the marketing wonks nearly as happy.

-- 
Daniel Morgan
University of Washington
Puget Sound Oracle Users Group