Kirk works for Kodewerk as a consultant offering Java performance-related services and training. He has developed and tuned applications written in C/C++, Smalltalk, and a variety of other languages. Kirk has written many articles and spoken at several conferences on the subject of performance tuning. He helped evolve http://www.javaperformancetuning.com as a resource for performance tuning tips and information. Recently he founded jClarity which is focused on building better performance diagnostic tooling.
After a brief introduction to a methodology to performance tune Java applications, the audience will guide me through the steps needed to tune an application. The application models performance problems that are common in real world applications. During the session, I will introduce a number of tools designed to expose causal code paths for each specific problem.
In Java 9, the G1GC will become the default collector and the mostly CMS collector will be deprecated. The hype is that the G1GC is suppose to be easier to tune. In this session we will explore the reality by looking at the performance implications of moving to the G1 collector. The session will draw upon experiences with tuning the collector in production environments. This isn’t simply about performance. A sub-optimally tuned collector will increase your companies operating cost by 10-20%. We will also explore results from benchmarks that expose potential pain points. and offer some suggestions on what actions can offer some relief.