Research On Network Simulation Partitioning Method On Heterogeneous Computing Environment

Network simulation, especially large-scale network simulation, typically requires enormous computing resources. Because of limited computing capacity of single simulator, the scale of network simulation can not meet the actual needs of research, therefore, current large-scale network simulation commonly uses parallel distributed network simulator. Specifically, first divide a large-scale network simulation tasks, then calculated by multi cluster nodes, thereby increase the scale of network simulation.However, the good and bad ways of partitioning simulated topology, often leads to huge differences in the efficiency of the simulation. at present, topology partition of parallel simulation usually uses topology graph partition, that is to induce actual network simulate task to a weighted graph, Then use certain tools to complete the partition of simulation tasks, such as METIS, Chaco; These traditional tools usually perform very good when the simulation environment is of the same structure. Generated topology often needs optimum time to complete. However, when the nodes of simulation platform is not the same as computing power, and the link between the nodes form a complex network topology. At this point if we do not consider the simulation platform topologies, only simple using the traditional topology partition tools, roughly assign network topology to be simulated to the nodes of the platform, then, the simulation procedure of the entire platform will be difficult to achieve efficiency optimal, resulting in computing power a tremendous waste.This article introduces a Cluster Instructed Partitioning Algorithm (CIPA), suitable for partitioning simulation tasks for parallel heterogeneous platform, greatly improved the efficiency of the entire system, and implemented an automated network topology partitioning tool, able to convert single simulation scripts to parallel simulation scripts through complex computing and evaluation. Compared with traditional semi-manual mode, greatly efficiency improved.