Re: A kind of... benchmark?

Rick Denoire <100.17706_at_germanynet.de> wrote in news:5rg5jvgsnhaofispfvbdotpbrtlbj4qruq_at_4ax.com:

> The simplest case. Run the benchmark. Reorganize the database. Rerun
> the benchmark. Kill a myth or stick at it.

Short of running a full blown TPC (it's not that difficult to setup a host-mode run, even a two-tier run), I'd suggest you capture your existing application's SQL (sproc's I hope!) during a 'typical' operational period and replay it. Also, I'd also ensure there are time stamps so you can replicate the 'think time' into your benchmark which is very important.

This will be the best benchmark environment for you rather than trying to use a TPC-C vendor kit. Even the kits requiring tuning.

When we've run TPC-C's, we take the toolkit from the vendor and give it a go around the block. We tune the box etc and then tune the kit. The kit is a good baseline to start but it's not The Best Configuration for all platforms.

-- 
Pablo Sanchez, Blueoak Database Engineering
http://www.blueoakdb.com