A dynamic grouping scheduling for heterogeneous Internet-centric metacomputing system

In this paper we propose a dynamic scheduling heuristic for allocating task graphs onto Internet-centric metacomputing systems. This proposed algorithm, which is called the Dynamic Grouping Scheduling (DGS), differs from conventional algorithms in three respects. First, DGS employs a dynamic task grouping strategy to determine task computational cast. Second, this algorithm estimates the competence of processors for unscheduled tasks at each scheduling step. Third, the proposed scheme takes the fluctuations of power weight for processors into consideration. Experimental results show that the proposed DGS performs better than the competing scheduling schemes under the effects of varying power weights of processors.