Estimating the performance of multimicrocomputer networks
A new paradigm for parallel computation based on large numbers of interconnected microcomputer nodes has recently emerged. The absence of any memory shared among the nodes makes the selection of an interconnection network capable of efficiently supporting message passing critical. A network performance evaluation technique based on the asymptotic analysis of a closed queueing network is used to compare six proposed interconnection networks: the single ring, the spanning bus hypercube, the dual bus hypercube, the toroid, the cube-connected cycles, and the r-ary m-cube. Among the questions considered for each network are the theoretical upper bounds on performance, the minimum sized computation quanta efficiently supported, and the number of nodes over which each network is preferred. 15 references.
- OSTI ID:
- 5169802
- Country of Publication:
- United States
- Language:
- English
Similar Records
Parallel computing on a hypercube: An overview of the architecture and some applications
Performance of the direct binary n-cube network for multiprocessors