Optimizing the performance of a J2EE application is a complex task, due to the nature of the workload which involves a network of connected computers in multiple tiers. The performance characteristics of the workload can be unexpected unless the statistics are adequately monitored. When the software or hardware systems are upgraded, the performance bottlenecks might shift from one place to another. It is important to apply a top-down, data-driven approach to identify and remove bottlenecks outside the focus of application-server performance. One can study and tune the performance on the application server only after the other bottlenecks are resolved.

Removing the bottlenecks outside the application server focuses on allocating resources so that those bottlenecks can be resolved. Optimizing the performance of the application server focuses on using the resources on the application server wisely so it can do more work. This article discussed data and tools to collect such data. There is usually too much data to look at it all, and it is unclear what data is critical in the early stage. Thus, getting comprehensive baseline data is a necessary starting point, so that future data can be compared against that baseline.

The same tools and methodology can be used to detect performance issues for different workloads. It is very important, however, to use tools that are as non-intrusive as possible, in order to provide a true analysis of the real environment. One simple way to improve performance of your application is to use a JVM that is already optimized for your platform. Understanding the JVM and additional tuning can yield additional benefits.