Performance Evaluation of Two-Dimensional Distributed Factoring Self-Scheduling Scheme for Heterogeneous Computer Systems

A major characteristic of distributed computing systems is their heterogeneity. In practical computing, the computers or processing elements that make up any given distributed system may have varying processing speeds. Efficient loop scheduling schemes for concurrent processing of computational tasks on such systems need to take into consideration, the varying speeds of the component processors. In this paper we evaluate the performance of a twodimensional distributed factoring self-scheduling scheme for parallel loops by comparing with its onedimensional counterpart. The schemes are implemented on the Ranger Computer Cluster at the Texas Advanced Computing Center with varying number of processors and problem sizes.