Is there any chance you can use a profiler on these tests? This should makes things a lot clearer.
If you're using Eclipse, there is an eclipse profiler that is free, easy to set up and very informative. http://sourceforge.net/projects/eclipsecolorer/ -----Original Message----- From: Vy Ho [mailto:[EMAIL PROTECTED] Sent: 10 January 2005 17:39 To: [EMAIL PROTECTED] Subject: Re: axis performance in multithreading env Using the latest source code (ant build from today's source download) (will be called "New code") and the Axis 1.2 RC2 (will be called "Old code"), I run the test with different parameters. The client side is 1 application with 1 or more threads. Each thread makes a number of requests to the server using a same stub returned from a locator for each thread. The results show that the new code somehow is slower than the old code. I really don't want to jump into any conclusion here. I just want to show this, and hopefully someone can get their benchmark on this too, and see why things turn out this way, or point out something I dind't do right. The average time is the sum of time each thread takes to finish and divides by the number of threads. Min and max time are not that far off from this average. New code: #threads #request Avg time (seconds) 200 50 116 100 100 110 5 2000 85 1 10000 138 Old code: #threads #request Avg time (seconds) 200 50 114 100 100 100 5 2000 81 1 10000 77
