Scheduling Byte Code-Defined Data Dependence Graphs of Object Oriented Programs

This paper presents pre-optimization methods for distributed Java applications on Java Virtual Machines (JVMs), which reduce program execution time. On the basis of an analysis the byte code compiled for a program, an extended macro data flow graph is generated. To this graph clustering and scheduling algorithms are applied based on modified DSC heuristics. The algorithms define distribution of clusterized nodes and objects on JVMs, so as to reduce the total execution time of programs. The nodes of the macro data flow graph are distributed to different JVMs as dynamic threads of distributed objects by modification of the program byte code without modifications of Java virtual machines.