Comments on blog.m1key.me: Scala performance - garbage collection analysis

Thanks for all your replies. @jigarshah.net You h...

2010-04-07T09:29:40.492+02:00

Thanks for all your replies.

@jigarshah.net
You have these parameters you can pass on to the server:
-XX:PermSize=
-XX:MaxPermSize=
As far as I know, the server is responsible for this, it has a memory leak of some sort. I found more information here:
http://www.jroller.com/agileanswers/entry/preventing_java_s_java_lang

"The message is a symptom of an incomplete garbage collection sweep where resources are not properly released upon unload/restart. [...] the finger of blame has been pointed at CGLIB, Hibernate, Tomcat, and even Sun's JVM."

jigarshah.net I am not sure mau be this can help y...

2010-04-07T08:46:57.257+02:00

jigarshah.net I am not sure mau be this can help you with PermGen problem, see http://www.tomcatexpert.com/blog/2010/04/06/tomcats-new-memory-leak-prevention-and-detection

Hi thanks for brief overview about the same. Enjoy...

2010-04-07T08:38:42.464+02:00

Hi thanks for brief overview about the same. Enjoyed it. I am facing another issue. Its PermGen space after redeployment. a well known issue on the web forums. Can you please suggest me how to look for details on the same ? Or can you point to some place Where i can get a little more help ?

Good garbage optimization is better then no garbag...

2010-04-06T21:10:37.539+02:00

Good garbage optimization is better then no garbage optimization.
Can't argue on that one :)

Vincent, to partially answer your question, you mi...

2010-04-04T14:32:00.938+02:00

Vincent, to partially answer your question, you might be interested in seeing the output generated by your program when the JVM is given the hprof option.

java -Xrunhprof[:options] your_program

It will generate a report for heap usage. If you ask it to generate a binary report (Xrunhprof:format=b), you can analyze it with Eclipse Memory Analyzer.

I ran this for my Scala program and I guess the result was similar to the one you predicted. Quite a lot of Scala garbage, it even detected two potential memory leaks.

Perhaps you'd like to try out this program yourself. It has quite a few options and you might find what you are looking for.
http://www.eclipse.org/mat/

I think this is a very interesting subject to stud...

2010-04-04T01:24:51.856+02:00

I think this is a very interesting subject to study, as more languages are created (and migrate) to the JVM, there may be vastly different amounts and types of garbage generated for each.

I wonder if there is an easy way to see the average object lifetime, and compare scala programs to java programs.

My thought is scala will generate more short term garbage, but I don't have any proof.

The G1 Collector has some interesting properties, and though it was originally intended as a 'concurrent' collector (which usually sacrifices overall througput), its certainly possible it could be even faster for scala like programs.

Hey Vincent, thanks for your comment. Well, if an...

2010-04-01T23:57:49.952+02:00

Hey Vincent, thanks for your comment.

Well, if an object lives a long life it will enter another generation (tenured). Then it is no longer subject to minor collections.

But you're right. If they all die young enough - then major collections will not run at all. I will update the article.

Best regards.

"the bigger the young generation, the less of...

2010-04-01T23:50:01.365+02:00

"the bigger the young generation, the less often minor collections happen - but the more often major collections will)"

I'm not sure this is true, I think it would depend on your average object lifetime.

If no object exists for more than a few milliseconds, why would there be more major collections? They should never make it to the old generation to begin with.