关键词: 调度, 启发式算法, 异构, 集群

Abstract: When vast data collected from broad-band network enter the parallel computing network，the data that are much feeble communication signals will realize the process of high gain and low delay and then achieve the purpose of explaining communication data efficiently by using cluster computing platform.The paper proposes a novel dynamic heuristic scheduling algorithm named MDS（Multi-Dimensional Scheduling） algorithm，which takes timing constraints，throughput and load balancing into account while scheduling.The MDS algorithm is capable of achieving high guarantee ratio even when the tasks have tight deadlines.In addition，the MDS algorithm can make the system have high throughput and achieve load balancing on the basis of high guarantee ratio.By several simulation experiments，this paper analyzes the performance impact of some parameters and compares the MDS algorithm with two best known algorithms.The experimental results show that the MDS algorithm outperforms other ones.

Key words: scheduling, heuristic algorithm, heterogeneous, cluster