In: LNCS 1823: High Performance Computing and Networking, pages 657-667. 8th International Conference, HPCN Europe 2000, Amsterdam, The Netherlands, May 2000. Proceedings / M. Bubak, H. Afsarmanesh, R. Williams, B. Hertzberger (Eds.) --- Springer Verlag, 2000.
Abstract: Understanding and improving Java performance is an important objective for both application and tool developers. Current efforts toward developing benchmarks have played a crucial role toward this objective, particularly in identifying general trends that can be observed through such a study. An alternative mechanism for categorizing Java performance is investigated here, based on developing a performance model of Java execution. The presented model is based on stochastic Petri nets and aimed at evaluation performance of method calls and class loading in Java programs. The primary objective of this paper is to use the Java profiler to determine distributions of execution time for method executions, and use this as a first step in analyzing Java performance using stochastic Petri net models.
Keywords: Java, performance analysis, stochastic Petri nets.