Hi David --
We do performance testing of the various JDK versions as well. In fact,
last week Sun posted new results on the SPECjAppServer 2004 benchmark
using GF 9.0 UR 1 and JDK 6. You can't actually tell directly from the
result, but in our testing using JDK 6 improved our score on that
benchmark by about 6-7% (which is of course why we submitted the results
on JDK 6).
You shouldn't actually need to recompile your application. However, if
recompilation were an issue, it would have shown up in the Jetty test as
well, so I mention this just to save you some trouble in the future.
I suppose it could be some issue with the AJP connector, which is
something we don't use in any of our performance tests. Do you see the
same issue if you access the appserver directly instead of going through
apache?
-Scott
On Tue, 2006-11-28 at 10:18, David Harrigan wrote:
> Hi Everyone,
>
> GF v2 b25.
>
> I've come across another performance issue, this time it manifests when I
> switch JDK. I've got the same machine and same setup as before, but this
> time I'm switching JDK's (since we hope to go to JDK6 Q1 2007). Anyway, I've
> performed the same load testing that I have always been doing and I'm
> getting some interesting (and worrying!) results.
>
> I'm testing this against JDK 1.5.0_09 and JDK 1.6 rc b104. Before each
> switch I rm -rf the domains directory and I run ant -f setup.xml to
> recreate. Then I add 3 jvm-options to the domain.xml
>
> com.sun.enterprise.web.connector.enableJK=8009
> com.sun.enterprise.server.ss.ASQuickStartup=false
> isContainer=true <-- this is an option for my application and not for GF.
>
> The application is rebuild each time, using the standarised build.xml that
> we here are using, but I'm switching source level to 1.6 when I switch JDK
> to version 1.6 (I put the source level to 1.5 when I switch back to JDK
> 1.5).
>
> I again get a dramatic slowdown running JDK 1.6. My average TPS is about
> 1.5-2, whereas in JDK 1.5 I'm hitting nearer 16tps.
>
> I thought, hmm, could be a JDK issue, so I deployed the same app in Jetty
> and then JRun and the TPS under JDK 1.6. is up around 16, just like running
> under JDK 1.5. (For completeness, I rebuild to source level 1.5 and ran with
> JDK 1.5 and the tps is again around 16).
>
> So, the only thing changing is the app server, i.e., GF ;)
>
> (oh, and I have no system.out's coming out :-) see passim for details on
> that one...)
>
> Has anyone else tested the same app between JDK versions? It could very well
> be a GF issue however.
>
> -=david=-
>
> --
> View this message in context: http://www.nabble.com/Another-performance-issue%2C-JDK5-vs-JDK6-tf2718901.html#a7580961
> Sent from the java.net - glassfish users mailing list archive at Nabble.com.
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: users-unsubscribe_at_glassfish.dev.java.net
> For additional commands, e-mail: users-help_at_glassfish.dev.java.net
>