Comparing JEE Servers
When picking which JEE server to use for your application, you have a number of choices to select from. Knowing which application server is the best is key. Recently Jonathan Campbell took a handful of JEE application servers, and came up with surprising results as well as informative comments.
It wasn't too long ago that Jonathan compared a handful of JEE application servers for Javaworld. The J2EE server being compared then were JBoss 4.2, Geronimo, and Tomcat 6. During this presentation the servers were being compared on features, deployment, and performance.
Java EE 5 compliance
EJB 3.0 capable
JSP 2.1 and Servlet 2.5 capable
Hibernate 3.x support
Along with a feature comparison Jonathon also did some performance benchmarking. Testing with simple JSP's and a servlet, tests were ran on the application servers to see how session creation handled as well as the speed of user connections while maintain concurrent sessions. Geronimo came out the winner based off of scores with having the most features. Overall, Jonathan concluded that any of these servers would be able to get the job done. The full list of the features being compared as well as how each of the JEE application server compared can be found here.
This time around, a different set of JEE application server where being compared. This time the contenders were Glassfish 2.0, JBoss 4.2, WebSphere 2.0 Community Edition (Geronimo), and Weblogic 10. The same features that were compared last time were also taken into consideration for this new set of JEE application servers.
From a sheer features standpoint, Geronimo was the winner, with Glassfish 2.1 coming in just a step behind. Performance benchmarking was done again on these server, but this time an application called jRealBench was used. JRealBench, designed to show realistic benchmarks, concentrates on testing session creation / hits as well as rehits. Geronimo came out ahead again in the benchmarks, with JBoss in second.
Once again Jonathon concluded that Geronimo was then winner when getting the most out of you dollar. That sparked a number of comments that brought to question the validity of the results. Some arguments were based off the idea that each JEE application server has its own tailored install / JVM such as Weblogic's JRockit, which provides additional performance that where not taken into account. Others mention SPEC, a non-profit corporation that has benchmarking JEE application servers for a while now.
In the end, Jonathan provided a straight out of the box, straight forward benchmarking comparison of some of the leading JEE application servers. The complete list of features compared and results can be found here.
Performance and Features
Different benchmarks measure different things. Scott Oaks, Sun's lead for our Java-server side efforts looked at Jonathan's post and wrote a response. See it here and here
Feature comparisons will depend on what's measured and how; what's in the table and what is not, and very subjective interpretations (like Usability).
I expect more from InfoQ
What does it brings to know that an IBM product wins an IBM partner test ?!
Next time please put a big clear warning for this kind of "news" !