JAVA FORK-JOIN PARALLEL GAUSS-SEIDEL IN AMAZON EC2 CLOUD

Abstract

From its 1st version, the Javalanguage supportsparallel programming using „ threads“, which are the part of Java language from the beginning. In this field, Java Standard Edition 7 (Java SE 7) brings new opportunities, like the new high-level framework Java Fork/Join, for simpler implementations of high-performance parallel programs. This paper presents an implementation of the parallel Gauss-Seidel algorithm, using of a modern programming language Java SE 7 and a new framework Java Fork/Join. Besides construncting the parallel Gauss-Seidel algorithm, the papers explains using the Amazon EC2 cloud for the execution of parallel Gauss-Seidel program. Performancecharacteristics of the parallel Gauss-Seidel algorithm have been easured, while the MATLAB software have been exploited to generatelinear sistems of different size and properties. Big linear sistems (dimension 10,000x10,000) have been tested, and good execution times obtained (less than 1 second) in Amazon EC2 cloud, which reccomends cloud computing forparallel processing of big linear sistems.