[concurrency-interest] Java concurrency benchmarks - need ideas

Krzysiek Zaporowski krzysiek.zaporowski at googlemail.com
Wed Nov 28 09:04:41 EST 2007


Hi,
In connection to my previous topic about ReentranLock benchmark - I'm
doing a little research about concurrency in Java for my University
and have to make series of benchmarks.
What is has to be done is a conclusions such like "yeah, ReentrantLock
is good for more scalable locking, but if you need simple lock a
non-nested block of code, it's better to use synchronized because of
less overhead in today versions of JDK 5 and 6".
In my benchmarks I'm trying to follow mostly JCIP book, doing for
example performance measurements of concurrent collection classes by
implementing for example producer-consumer pattern with various
consumer load of work degree, various consumer threads number, etc.
I've also measured overhead of locking (my previous post). I will also
measure mean Thread creation and start() time taken.
But I don't have idea for simple use-cases of concurrency, that will
lead me to make conclusions like above. Have you got any ideas what
should be measured?


More information about the Concurrency-interest mailing list