Tapestry 5 Benchmarking - Update

Following my previous post on Tapestry 5 benchmarking I am now going to expand the set of tests as suggested

Test Various JVM/OS Combinations

• Run the test on a dual CPU machine with
• Linux 32/Java 6
• Linux 64/Java 6
• Open Solaris 64/Java 6 32
• Open Solaris 64/Java 6 64
• Each test will be run for 1 minute. The first 20 results will be ignored. 40 samples (1 second)
• I did some 'pre-tests' for this and found the results varied (even with VMWare 'reserving' CPU's) so I will run each combination 5 times.

Test a Struts Application of similar complexity

• I have created a struts/tiles application that mirrors as closely as possible the Tapestry sample application
• Re-run the tests on both this and the original sample

Test absolute break condition

• In the first test I saw the site slow down under load but never break. This is unusual. Most J2EE sites I have seen hit a load point and break totally (usually filling a thread pool).
• I will run the test will more threads (on dual CPU) and see if I can find the break condition

Test a more complex tapestry / struts applications (if I have time)

• I will add a form on both the the applications with 2 fields and server side validation