| In recent years,with the wide application of mobile Internet,the rise of e-commerce and the rapid development of data services,the number of Internet users increases exponentially with the increase of the scale of Internet traffic,showing explosive growth mode at the same time,the extensive application of vitalization technology,cloud computing technology has developed rapidly,the traditional network architecture to TCP/IP as the core has been unable to meet the requirement of modern network,so it is difficult to maintain and expand the difficult problems have become increasingly prominent.These problems have limited the development prospects of the Internet and prompted people to study the new network architecture.Based on this,Software Defined Networking(SDN)was created.SDN differs from traditional network in that it uses layered thinking to separate data layer and control layer,thus realizing flexible and effective control of network traffic.However,a very important problem that SDN needs to solve is the load balancing problem of the network,It is closely related to the stability of the entire network and also seriously affect the network operation cost,so the quality of the network load balance is crucial.In the current network environment,the new network architecture(SDN)appear is an inevitable trend,then the load balancing technology of network research,we can be placed in the network architecture,which is of great significance for the innovation and development of the network.The purpose of this paper is to realize the load balancing of link under SDN architecture.At first,this paper proposes a traffic prediction algorithm based on traffic detection,the basic idea of the algorithm is based on detecting changes in the flow,the flow into the stability of the elephant flows and unstable flow of mice,on mice flow,uses the ECMP(Equal-Cost Multi-path Routing)algorithm for forwarding;On elephant flow,this paper then puts forward an improved IPSO(Improved Particles Swarm Optimization)algorithm,by improving the algorithm to find a optimal fixed path to fast forward steadily elephant flows,so as to avoid the elephant flows in centralized,in order to achieve the purpose of balancing network load.Finally,set up the SDN data center platform were built by network simulation software Mininet and Floodlight controller,the formation of custom network topology,then the single path,PSO and IPSO experiments and test the load balancing scheme,resulting in a large number of experimental data.Then,we collected the experimental data analysis and comparison and draw relevant conclusions.The results show that the IPSO algorithm has better reliability,better able to achieve the load balance of the network link. |
