Research On Background Traffic Modeling And Generation Method Based On Traffic Matrix

With the development of science and technology, the scale of the Internet has been expanding. Optimizing network structure, monitoring network behavior and design of high-performance network protocols have been paid more and more attention by researchers. Network research has become one of the hot topics in the field of computer science. Researchers usually use three methods that analysis, experiment and Simulation in the study of network. No matter which method is used in network research, it only has traffic of testing the application or new protocol traffic and has no other traffic in the new experimental network environment, which with real network environment have very difference. Therefore, network application or new protocol is working well in the simulation of the network environment, but not well in the real network environment. In order to make the network environment more close to the real network environment, we need to add the background traffic to interference foreground traffic in simulation or emulation of the network environment. The similarity degree between the spatial and temporal characteristics of the generated traffic and the real network traffic affects the efficiency of the network application or network protocol development. How to generate more realistic background traffic has become a hot research topic.Some traffic models of existed traffic model only pay attention to the generated traffic whether keep the temporal characteristics of the real traffic and other models of existed traffic model focus only on generated traffic whether retain the spatial properties of the real traffic. In view of this situation, In this paper proposed a kind of background traffic modeling and generation method based on the traffic matrix. This method preserves the spatial and temporal characteristics of the real traffic.We present the background traffic modeling and generation method based on traffic matrix is mainly divided into three steps. First step we analysis and cluster the original network traffic. we treat the unique IP address as end host and extract four attributes of each of IP addresses. Then the K-means clustering algorithm is used to cluster the IP address, and we collect the statistics of the traffic flows between the cluster. The second step, we use integer linear programming method according to router information of a given network topology to obtain the traffic matrix to correlate the clustering results and routers of the given network topology. The third step is to use the result of the first step and the second step to generate background traffic.In order to verify the effectiveness of our background traffic model, we use the rand index to examine whether the generated background traffic would constitute the same type of flows as in the original traffic. Because the Abilene network has a real traffic matrix, we use the information of Abilene network compare with our background traffic model to validate spatial distribution and temporal characteristics of background traffic of this paper presented.