In practical performance tests using postal, I have not seen a measurable
improvement using the Hotspot server.
The use of object pooling in James v2.1 significantly reduces object
creation. The JVM heap tools indicate that James is about as clean as it
can be in terms of object creation. I have a revised model in mind for the
protocol handlers to address both socket and nio based I/O, which would
reduce much of the current overhead from I/O wrappers, but that's a story
for another month.
Again, YMMV. I can only report the measurements I've made. And I have not
tried turning the heap generation parameters as yet.
--- Noel
--
To unsubscribe, e-mail: <mailto:[EMAIL PROTECTED]>
For additional commands, e-mail: <mailto:[EMAIL PROTECTED]>
