From: Robert Klemme <shortcutter@...>
Date: 2006-08-08T17:15:14+09:00
Subject: Re: Standard dataset for Database performance testing

M. Edward (Ed) Borasky wrote:
> Jon Egil Strand wrote:
>> Greetings
>>
>> Does there in the db-world exist any de-facto standard dataset for 
>> db-performance testing?
>>
>> I know there esists some "standard" problems both in image-analysis (some
>> picture of a woman, can't remember her name just now), and machine
>> learning (us zip-code hand writing). Are there any similar sets for
>> databases?
>>
>> Ideally, it should be large, contain multiple datatypes, some structure
>> but still be easy enough to present in a talk.
>>
>> Reason:
>>
>> I'm planning on doing some db-performance testing.
>>
>> Goals:   a) Learn more about Ruby's db tools
>>   b) Learn more about postgres & sqlserver (my main options at work)
>>   c) Learn profiling and tuning
>>
>>
>> All the best
>>
>>   
> There is indeed a "standard" database test suite. I believe the base 
> implementations are in C with PostgreSQL as the database, but of course 
> you can vary the language and the database for purposes of comparison. 
> The suite can be found at
> 
> http://sourceforge.net/projects/osdldbt

Hm, didn't know that.  Thanks for that link!

However, I'm not sure about the usefulness of such a thing.  First, to 
get meaningful results the dataset typically needs to be huge - but 
there is huge and huge (for some apps it's GB for others TB).  Second, 
each application's DB schema and usage pattern differ, so while it 
probably makes sense to have a SAP DB test suite I'm not sure what I'd 
expect from a generalized test set.  Did you use this already and can 
share some experience?

Kind regards

	robert