Web Performance, Inc. has published a performance report comparing tests of Apache Tomcat on Windows and Linux, with interesting results. The report found that Linux was able to handle about 32% more users than Windows on identical hardware with identical test conditions.
What I love about this report is the level of detail they provide about their methodology and the data. They provide all the information needed to duplicate their results. From the press release:
No matter how this testing is performed, someone will scream “Foul Play!" A primary goal of this effort is a test that is easily duplicated by anyone else who wants to try. No doubt the exact numbers will not be duplicated, due to differences in hardware and measurement technique, but the relative differences between the servers should be reproducible. All of the materials related to this report in the Supplemental Materials section.
Anyone interested in performance benchmarks of Java-related technologies on Linux and Windows should definitely check out the report.
The company also published an earlier performance report on Java servlet containers. Note the interesting commentary (in the section titled “Servers") about how some proprietary licenses block the publication of benchmark data without permission. Even though they conducted the tests and had the data ready, they were never given permission to publish. What do the proprietary companies have to hide?