Heterogeneous clustering computing based on parallel task distribution

Abstract

To improve the computing efficiency of heterogeneous cluster, this paper presents a parallel task distribution approach for heterogeneous cluster. According to the computing power of each node in the cluster, the paper computes the weight of each node dynamically from the perspective of the predicated evaluation of system pressure. Then, the sub-task in each task period is distributed for each computing node in accordance with its weight, which can insure that each node can be assigned a rationale task and thus make a global optimum for the cluster. Experimental results demonstrated that the approach presented in this paper can guarantee the effective load of computing nodes and improve the service quality of cluster as well. © Springer-Verlag Berlin Heidelberg 2012.