Cooperative Communication System for Parallel Computing Agents in Multi-cluster Grid

The idle computational resources of CSCW environment that is composed of computer clusters are mined to construct the multi-cluster grid in order to support the computation-intensive tasks. For fitting the state changes of idle computing resources during the computing process, the techniques of cooperation and migration agents are adopted. Through the concurrent dataflow, the supper element, and the density factor, the dynamic buffer pool (DBP) was built up. By using of the Grid techniques, the cooperative computing agent, the cooperative communication agent, and DBP, a cooperative communication system (CCMS) for parallel computing agent was designed and implemented. The experimental results show that CCMS can increase the speed-up of parallel computing task. It can be fit for the computation in CSCW based on Internet.