I know that my tests don't compare directly to the other post, and it doesn't look like there is as much of a change in performance as the other tests showed between versions. The HelloWorldExample servlet was 1.2% faster using 5.0.19 compared to 4.1.30 with the HTTP Connector. It was 7.2% faster when comparing 5.0.19 and 4.1.30 using the jk2 Connector.
There's no difference because the HWE is a rather bad test (if you're out there to test throughtput):
- it retrieves an i18n bundle on each request, and gets a String from it (profiling has shown it's not something trivial)
- it uses a writer: this means most of the activity is converting the chars to bytes
Since it doesn't exercise anything else (filters, request dispatcher, any API method), there's little difference between 4.1.x and 5.0.x (which now also use the same connectors, so there's no difference there either).
If you want to test the output speed, you need a servlet which simply writes byte arrays (and does nothing else).
You could also use keepalive when testing (unfortunately, ab doesn't allow a "mix"). Again, you'll see bigger differences.
-- xxxxxxxxxxxxxxxxxxxxxxxxx R�my Maucherat Developer & Consultant JBoss Group (Europe) S�RL xxxxxxxxxxxxxxxxxxxxxxxxx
--------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
