3.2.2 performance tests

[fbiaggi]

Hi,
after some tests, the 3.2.2 release, in an EJB environment show 22% better performance.
Very good.

[jcordes]

Hi !

I've experienced performance gains, too (about 25%). May be a result of the massive code-cleanups ;-). More performanced can be gained by using jrockit or jikes with sun jdk (nearly no improvements by using jikes with jrockit).

Very, very good. I'm investigating for more tuning possibilities :-).

[lafr]

Using prepared-Statements (added <prepared-statement-cache-size>1000</prepared-statement-cache-size> to the datasource config) doubled performance for my test app.

[jonlee]

It is faster with the tweaking that has been done. However, I notice that 3.2.2 seems more sensitive to lock-ups during deployments. While I don't get any slow downs with repeated deployments I noticed that occassionally things would get tangled and JBoss would stop responding - hung partway through the undeply/deploy cycle and wouldn't respond to a shutdown.

I have also noticed the lock up problem with MBean deployments. I was changing an MBean structure and unfortunately forgot to update the deployment descriptors. When it got to the deployment on a clean boot, JBoss hung at that stage. In normal operation though, there are no problems.

Just things of which to be aware.

[genman]

If you can get a solid stack trace (kill -3 PID) of the JBoss process and file a bug, that'd be swell.. I haven't seen the same problems you describe.

[jonlee]

I'll see what I can do to provoke a lockup again. I'm just in the process of rationalising resource links and lookups with XDoclet.

[parikhm]

I do not see preparedstatement in the dtd for the datasource? where do I specify the prepared statement cache?