db-derby-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Paulo Jesus" <paulsch...@gmail.com>
Subject Re: Question on benchmarks on derby.
Date Mon, 03 Sep 2007 16:47:44 GMT
2007/9/3, Bryan Pendleton <bpendleton@amberpoint.com>:
> > We have tried TPC-W on Derby 10.2.2(server mode) and was not a viable teste.
> Have you tried Derby 10.3, which was just released? It would be interesting
> to know how the results compared.

My search is for a project developed in Derby 10.2.2. It was not
ported yet for 10.3.
The project has code embedded in the server engine.  I have follow the
developing of 10.3.

> > This test have many IN  statements with multilevel querys and max rows
> > statements. The optimizer doesn't make the best choices, there are
> > many table scans over indexed columns.  And have some memory leak. So
> I'm sure that the developers list would be quite interested in
> discussing the problems that you found, and the possible solutions
> that you think would be useful. Sounds like you may be hitting some
> known problems, but may also have encountered some new ones.
When i saw  the problems with IN statements i had some help from Army here.
Bad decisions from optimizer can be reduced by rewriting sub querys
from multilevel querys to FROM column (making temporary tables with
sub querys).  I found one paper with these. "Adding Subqueries
toMySql, or what Does it Take to Have a Decision-Support Engine?,
Chanda, M. and Badia, A.  Adding, Technical Report 02-01."
Is not from Derby, i know, but useful.

> What is your goal with respect to running these benchmarks? Are you
> attempting to evaluate whether Derby's performance would be acceptable
> for your environment? Or are you investigating Derby performance
> issues in order to address them?

I have been investigating Derby performance issues. And now for testing
the correction of engine, after patches being inserted and verifying
performance again.

> thanks,
> bryan

View raw message