The effect of task scheduling is one of the important factors which decide the performance of the parallel distributed computer system. In order to optimize the task scheduling, this paper proposed a novel algorithm based on some typical algorithms e. g. LG, PPA algorithm etc. . On the one hand, the algorithm replicated precursor tasks which meet the duplication conditions in order to shorten the scheduling length. On the other hand, the algorithm replicated other precursor tasks and merged redundant clusters to reduce the number of the required processors. Simulation results show that the proposed algorithm's performance on scheduling length and the number of the required processors is better than the above typical algorithms'. Furthermore, the proposed algorithm has a smaller time complexity and benefits the performance of the parallel distributed computer system.