Research On Efficient Network Simulation Method Based On NS3

Network simulation technology is a method of using mathematical modeling and statistical methods to simulate network behavior,abstracting network devices and network links in the real network environment,and simulating the traffic interaction behavior between network devices in the simulation environment.The research on network behavior and network structure is of great significance.Among the existing network simulation systems,the NS3 simulation system has gradually become the mainstream network simulation system due to its open-source nature and powerful module library.Although the NS3 simulation system can provide support for large-scale network parallel simulation,it still has optimizable content.At present,the NS3 simulation system can add simulation traffic to the simulation network.Users can set parameters such as the start and end time,traffic rate,and message size for a single traffic.However,the NS3 simulation system does not provide a solution to ensure the similarity between simulation traffic and actual traffic.At the same time,although the NS3 simulation system supports parallel simulation to perform network simulation,users need to specify the process numbers of different nodes in the simulation network before simulation,the NS3 simulation system does not provide a suitable node process allocation scheme to ensure the simulation performance of parallel simulation.This article optimizes the above two aspects,and the specific content is as follows:(1)This thesis presents a method of modeling and generating process simulation traffic based on Markov process.In order to reflect the real traffic at a certain time and in a certain scenario,first collect and obtain sampling traffic from the real environment.After analyzing the sampling traffic,obtain the packet size and state partition of the packet interval.Use statistical methods to estimate the transition probability between different states.Finally,combine the random number generator to generate simulation traffic,and verify through experiments that the generated simulation traffic has good similarity with the sampling traffic.(2)This thesis proposes a parallel simulation optimization method that combines graph partitioning algorithm.Assigning process numbers to nodes in a network topology before parallel simulation essentially involves dividing the network topology into regions.This article first studies the main factors that affect the performance of network simulation,and then uses the shortest path algorithm to predict the point and edge weights of the network topology.Finally,the graph partitioning algorithm is combined to partition the network topology into regions,and the optimization effect of the graph partitioning algorithm on parallel simulation is verified through experiments.