Load balancing for regular data-parallel applications on workstation network
A cluster of machines connected by a high-speed interconnection network is emerging as a new architecture for high-performance computing. Among the important issues that need to be addressed in this type of computing environment are adaptive load balancing and data partitioning. In this paper we discuss the parallelization of matrix multiplication and Gaussian elimination on adaptive and nonuniform environments. We present a simple strategy to reduce communication cost when remapping the arrays of these applications. We also develop CYCLIC distribution scheme for adaptive and nonuniform environments. Finally, we present performance results for the solution of the two applications on a cluster of heterogeneous workstations.
Unable to display preview. Download preview PDF.
- 1.Edjlali, G., Agrawal, G., Sussman, A., Saltz, J.: Data-Parallel Programming in An Adaptive Environment. Proceedings of International Parallel Processing Symposium. April (1995) 827–832.Google Scholar
- 2.Nedeljkovic, N., Quinn, M.: Data-Parallel Programming on a Network of Heterogeneous Workstations. Proceedings of the First International Symposium on High-Performance Distributed Computing. September (1992) 28–36Google Scholar
- 3.Kaddoura, M., Ranka, S., Wang, A.: Array Decompositions for Non-Uniform Computational Environments. Journal of Parallel and Distributed Computing. 36 (1996) 91–105.Google Scholar
- 4.kaddoura, M., Ranka, S.: Run-time Support for Parallelization of Data-Parallel Applications on Adaptive and Non-Uniform Computational Environments. Proceedings of High Performance Distributed Computing. August (1996) 30–39.Google Scholar
- 5.Keyser, J., Lust, K., Roose, D.: Run-Time Load Balancing Support for Parallel Multiblock Euler/Navier-Stokes Code with Adaptive Refinement on Distributed Memory Computers. Parallel Computing. 20 (1994) 1069–1088.Google Scholar
- 6.Siegell, B., Steenkiste, P.: Automatic Generation of Parallel Programs with Dynamic Load Balancing. Proceedings of the Third International Symposium on High-Performance Distributed Computing. August (1994) 166–175Google Scholar