On Tue, Jan 6, 2009 at 1:16 AM, Pascal Cohen <pcohen@wimba.com> wrote:
> Hello and best wishes for this new year.
> I have a question concerning the H2 DB.
> http://www.h2database.com/html/main.html
> I've read (on their site) that they got better perfs than PG or MySQL in any
> case (embedded in a Java application and even as a standalone server).
> Tests seem a bit "light" with a single thread benchmarking the DB but the
> results seem anyway interesting.
So, in other words, it's not really that interesting. :) How well a
db runs with a single thread really doesn't mean a lot unless you're
only using it for single user embedded or batch processing. I'd like
to see a simple pgbench style (i.e. mixed reads and writes with
transactions) benchmark with 5, 10, 50, 100 users, etc... That would
tell you something interesting. Since they haven't published a
benchmark with > 1 user, I'm willing to bet that the performance with
> 1 users is not so good, and gets worse as you add users.
http://tweakers.net/reviews/649/8/database-test-sun-ultrasparc-t1-vs-punt-amd-opteron-pagina-8.html
Now that is an interesting benchmark. Notice how MySQL is a good
20-30% faster with one user? More importantly see what it does with
many users, and how it behaves as the number of users increases.
