On Oct 17, 2007, at 6:25 AM, Jan Kriesten wrote:
> a recent comparison between resin 3.0.9 and tomcat 5.5.3 can be
> found here:
> on this comparison maybe scott could comment?
It's not a useful benchmark. It measures startup time and JSP
compilation time, not runtime. I'm actually embarrassed for the
industry that Java people link to it as if it were a real
benchmark. (That's not to say startup time isn't important,
especially during development, but it's not a performance comparison.)
The problem with benchmarks currently is that most applications have
tons of stuff piled on top of the app server, e.g. spring, struts,
jsf, hibernate. So any particular benchmark might be measuring a
totally different environment than what you're running.
For example, we can run "hello, world" JSP on our load testing and
get something like 8000 requests per second. Add a basic JSF page
and you get 400 requests per second. And that's without using JPA or
Hibernate or application code.
Now that Resin has JPA and JSF, it may be useful to start
benchmarking again. If an application uses the core servlet engine,
JPA, and JSF, the time spent in Resin code is fairly significant, so
benchmark numbers would give meaningful results.
> regards, --- jan.
> resin-interest mailing list
resin-interest mailing list