Continuance Parallel Computation Grid Composed of Multi-Clusters

Abstract

For supporting the grid computing in dynamic network environment composed of multi-clusters, a continuance parallel computation grid (CPCG) is proposed in this paper. A series of formal definitions, such as the CPCG architecture, the dynamic network environment (DNE), the management agent system, the independent computing agents (ICA) which support the traditional computing (TC), the cooperation computing team (CCT) which supports the data parallel computing (DPC), and their relations are given. Through DPC, TC, and the migration mechanism, the continuance data parallel computing (CDPC) was constructed. The dynamic learning method, the fuzzy partition technique for the logical computer cluster on which CCT runs, the stage checkpoint mechanism and the migration process are studied. CPCG computing process is described. The experiment results show that CPCG resolves effectively the problems of optimization use of resources in DNE. It can be fit for grid computing.