XHTML Performance

[monkeyden]

I have been profiling our Seam 1.2 (XHTML) application using JProfiler and have discovered an obscene number of HashMap$Entry objects on the heap (typically about 7-8 million). Not withstanding the possibility that this is the most common data structure in a web container/application, I suspect that much of it comes from XHTML. I say this because I can easily reproduce it in the Seam Blog example (XHTML) but cannot in the Seam Messages example (JSP).

We're hosting on 64bit so we get more than 3GB of heap, however, during peak times transactions begin to timeout and the application fails to respond. It recovers for a short time but the threshold for recurrence is much lower.

I'm sure there are ways we can tweak our existing middleware to increase performance and decrease the footprint but this seems to be the most acute issue we have right now.

Can anyone share experiences or suggestions on this topic or is it the cost of using XHTML?

Thanks

[techieexchange]

Hi,
I would also like to hear experiences about this issue.
Because we also got some heap problems when running Seam+Facelets based applications.

http://seamexpress.wordpress.com/

[pmuir]

JBoss QA is going to be looking at performance testing and benchmarking for Seam over the next few months. We should be able to take the results from this to both publish a tutorial on performance and make improvements to Seam as possible.

[mgrouch]

JBoss QA is going to be looking at performance testing and benchmarking for Seam over the next few months. We should be able to take the results from this to both publish a tutorial on performance and make improvements to Seam as possible.

Are there any results available?
Thanks

[pmuir]

I was speaking to Jay about this on Friday. He's going to start looking performance analysis and scalability in the next few weeks. We've started a performance section:

http://alpha.seamframework.org:9999/Documentation/Performance