Distributed Java Programs Initial Mapping Based on Extremal Optimization

The motivation for the research reported in the paper is optimized execution of distributed Java programs in clusters of Java Virtual Machines (JVM) implemented on workstations and desktop Grids. Optimization of the execution time of distributed object programs has been always a challenging task due to specific execution paradigm of object programs and particular architectural features of the JVM. Several research papers have been written on dynamic load balancing and scheduling in clusters and Grids. In the cited solutions load balancing is done at Java program runtime based on centralized or distributed load monitoring agents. A good review of the works on task scheduling on Grids is given in. Efficient load balancing on Grid platforms requires adequate computational and communication load metrics.