Ole,
Let's not do the finger pointing. I trust you know what you are doing. We
also need a straighter test harness... so that commits don't result in
aberations like this one.
Hugo, if you test the JBoss beans by themselves can you get a time?
marc
|-----Original Message-----
|From: [EMAIL PROTECTED]
|[mailto:[EMAIL PROTECTED]]On Behalf Of Ole Husgaard
|Sent: Saturday, February 10, 2001 12:09 PM
|To: jBoss Developer
|Subject: Re: [jBoss-Dev] MAJOR SLOWDOWN in jboss' performance in a
|couple ofdays...
|
|
|Hi,
|
|I suspect some of this has to do with the
|changes to the embedded Tomcat, and some
|upcoming deployer changes.
|Probably calls are becoming non-optimized
|due to some classloader issue that is to
|be fixed in new deployer.
|But I don't use Tomcat, so I am not sure.
|
|For the JTA changes, these seem to me to
|have the same performance as before.
|These changes do touch the invocation
|layer, but not in any way that should
|result in optimized calls becoming non-
|optimized, or result in performance
|degradation.
|
|But the performance degradation you show
|below for after my changes is heavy, and
|I cannot say for sure that it has nothing
|to do with my changes.
|
|Depending on how much you checked out
|from CVS, it could also be due to
|dewayne's commit to module tomcat a few
|hours later.
|
|Unless someone knows something about
|what is happening here, I guess I have
|to learn to set up an embedded Tomcat so
|I can see for myself.
|
|
|Best Regards,
|
|Ole Husgaard.
|
|
|Hugo Josi Pinto wrote:
|>
|> Guys,
|>
|> I noticied the strangest behaviour while running the
|EmbeddedTomcat example.
|> In JBoss 2.0 final, I get these timmings, that look quite fine:
|>
|> 2.0 FINAL
|> Non-Optimized: Avg. time/call(ms):3.185
|> Optimized: Avg. time/call(ms):0.821
|>
|> In JBoss 09-Feb-2001 02:00 cvs snapshot, things look strange: results for
|> the optimized version of the test seem to be pretty slow and close to the
|> non-optimized one... like this:
|>
|> PRE2.1 09-Feb-2001 02:00
|> Non-Optimized: Avg. time/call(ms):1.893
|> Optimized: Avg. time/call(ms):1.452
|>
|> AND in today's cvs checkout, right after Ole's new TX/JTA
|changes, these are
|> the results:
|>
|> PRE2.1 CVS
|> Non-Optimized: Avg. time/call(ms):78.603
|> Optimized: Avg. time/call(ms):2.503
|>
|> WHOA! What happened there? 78 ms/ PER CALL? I don't belive these are good
|> metrics to go out to V2.1... there must be something wrong... I
|dont have a
|> profiler tool here at home, but you can get the .EAR's I was trying to
|> deploy at:
|>
|> http://www.hugopinto.com/files/jboss-ears.zip
|>
|> ...and try it out yourselves... Inside you'll find 2 files: the
|original 2.0
|> .ear, and a modified one, in which i inserted the Interceptor Chains into
|> the jboss.xml file.
|>
|> Regards,
|>
|> Hugo Pinto
|
