基于网格任务调度的Monte Carlo仿真建模

网格系统是异构和动态分布系统,其节点的计算能力存在差异。在多个节点共同完成任务时,容易产生计算瓶颈,影响计算吞吐量和造成延迟,甚至导致整个任务失败。利用Monte Carlo仿真的自然并行性,解决了网格任务调度问题,设计了基于网格的Monte Carlo仿真流程框架和节点状态机模型。在此基础上,建立了“N-M”调度策略的二项式概率模型,并在不同的参数条件下对Monte Carlo仿真任务的性能进行了测试和分析,结果验证了设计方案的有效性。

Monte Carlo Simulation Modeling for Grid-Based Task Schedule

Grid system is a heterogeneous and dynamic distributed system,its computing abilities are various for different Grid nodes.Bottleneck may come into being when carrying out task with several nodes,which will affect computing throughput and result in delay,even lead to total task failure.To solve the problem of Grid schedule,statistical nature of Monte Carlo simulation is utilized.Grid based Monte Carlo simulation framework and node state machine were designed.Based on that,A binomial probability model in terms of "N-M" schedule policy was established.Simulation performance under different parameter was also tested and analyzed.The result shows that design scenario is effective.