Abstract

Recently, large-scale computations have been used in many real application areas. At the same time, there exists more parallel computing technique on cluster in order to meet these demands. In this paper, a speedup analysis of Master-Slave application on heterogeneous cluster is investigated. A task allocation model is set up and its theoretical analysis of execution time is developed. A more accurate upper bound of speedup is derived under some conditions by virtue of the task allocation model. Furthermore, this theoretical analysis is verified by a group of experiments and the experimental results show that the speedup increases nonlinearly and rapidly which is caused by the task optimization in the process.

Key words: Master-Salve, speedup, cluster computing, heterogeneous cluster, super linear speedup.