In my article, I bemoaned the fact that it is hard to find high-quality experimental data that compares the performance of J2EE vs. .NET, as well as the performance of various J2EE servers.
A big step in the right direction has been provided by the Software Architectures and Component Technologies (SACT) group (http://www.cmis.csiro.au/SACT) that is part of CSIRO, an Australian research organization. This group has produced some detailed AND scientific investigations of the performance of various J2EE servers. They provided me with a complimentary copy of their latest report. It is not free, but it would be worth the investment if you have already committed to using J2EE and you are evaluating your options. Please do the industry a favor and support the efforts of unbiased groups like CSIRO! ;^)