Mesh Partitioning for Efficient Use of Distributed Systems
IEEE Transactions on Parallel and Distributed Systems
| Hi-index | 0.00 |
The improving computational speed of workstations makes clusters of these machines attractive for parallel computing. The workstation model of parallel processing, however, presents specific challenges caused by the latency of the communications network and the workload imbalance that arises from the heterogeneity of the nodes. Because of these issues, data partitioning is critically important for parallel processing on workstation clusters. In this paper, we mathematically characterize the communication cost for four data decomposition schemes: scatter, contiguous point, contiguous row, and block. These methods are analyzed in terms of problem size, number of processors, network speed, communication pattern, and bookkeeping requirements. Bounds are established for the performance of these decomposition schemes that can be used to make better-informed data partitioning decisions. Key Words: Data decomposition, workstation clusters, parallel processing, heterogeneous networks, interprocessor communication.