Dynamic Load Balancing Strategies in MPI Parallel Environment
Jiaxin Yu - · Computer Technology and Development · 2010
Cluster parallel computing environment is used more and more widely.MPI is the most important programming tool.In parallel processing,load balancing plays a very important role,it directly affects the efficiency of the entire algorithm.In this paper,under MPI programming environment specific characteristics,and based on the benefits of the method of valuation to determine whether trasfer to the task,given the real-time monitoring of load and scheduling algorithms.In each node interval test and compute.Finally,in the MPI environment,the use of parallel sequencing methods to verify the scheduling algorithms.The experimental results show that the load before and after the improvement has been obvious,especially in the larger task,result is more obvious.