By: Bill Buros.
Java based applications and workloads are pervasive in the server space. Nicely, IBM provides Power-tuned Java engines for the operating systems which run on Power servers. Easy to download. Easy to install. Easy to leverage. This allows product teams around the world to leverage the best of Java with the best of Power7 servers to deliver applications optimized for Power. IBM teams regularly provide the latest Java updates, with service, optimization, and technology updates.
Recently, a "Java performance best practices
" article was published with a suite of recommendations for tuning Linux applications running on Power7 servers. The article goes beyond the obvious steps of processor core affinity (which maximizes memory locality), but also helps to explain other system tuning considerations, scaling characteristics of JVMs with respect to number of processor cores, and a number of common Java-specific tuning options. The article is a handy checklist of things to consider in and around Java performance.
Leveraging a dynamic community, the Java article is being extended with new information being evaluated across a number of new workloads. We encourage community members to comment on the practices, try new areas of Java applications, and suggest additional tunings to be added. In the background, work is progressing on leveraging "perf" based tools, CPI breakdown analysis, and memory locality analysis to the tunings.