Research And Implementation Of The Distributed Parallel Network Simulation System Based On NS2

Network simulation is able to predict the validity and performance of the new protocols, applications, services and control algorithms before deploying them to the Internet, which makes it an efficient method to support the research and development of network related technologies. Currently there are two commonly used network simulation softwares, i.e., NS2 and OPNET.This paper aims to carry out intensive study on the distributed parallel NS2 system and cluster based job management system. The status quo of network simulation and the advantage of using NS2 are firstly introduced. At the same time, the problems happened when doing large scale network simulations using NS2 are put forward. Based on the survey, we found that enough RAM and CPU time are needed in NS2 when doing large scale network simulations, which makes the simulation very difficult in current status. To solve this problem, this paper sufficiently utilizes the adavantages of the grid computation and cluster system, and designs one PBS based job scheduling system to do the management and job scheduling of network simulation tasks. In this way, network simulations jobs can be done distributly and the resources of the cluter can be utilized adequately. Meanwhile, PDNS is adopted to decompose a single network simulation task into several parts, which makes a single network simulation task can be distributed to several work stations connected by TCP/IP. By doing so, the requirement of the RAM in single station can be reduced and the whole simulation time is reduced simultaneously.Finally, the paper concludes the work done during the Master study. Meanwhile, the possible improvements of this system as the future future work are also pointed out.