The Research And Design Of SDN Load Balancing Award Routing Scheme Based On Learning

With the rapid development of various new technologies on the Internet,the number of network users is increasing day by day,and the shortcomings of traditional network architectures that are overly bloated and poor in business sensitivity are increasingly exposed.Software Defined Networking(SDN)is the future network architecture.It has the new advantages of centralized control,open interface,and network programmable.The corresponding technology is the academic research hotspot.Among them,the research of load balancing technology in SDN environment has very important significance in the stability and cost of the network.This thesis takes the SDN three-layer network architecture defined by ONF,and the OpenFlow protocol to study the routing strategy of network links.First,we introduce the previous papers about load balancing algorithm in SDN network and classification algorithm applying to network traffic.Afterwards,the advantage of using machine learning to classify network traffic is introduced,and we propose a scheme of using machine learning classification method to solve the load balancing problem in SDN network.According to the different QoS requirements of different QoS applications and the characteristics of the traffic,the classification algorithm can classify the network traffic.Therefore,it can be obtained that different types of flows are assigned different paths when the client and the server are the same,thereby achieving the purpose of load balancing in SDN network.In this paper,the design of classification algorithm and the design of the routing algorithm to implement network load balancing are introduced in detail.We verify the algorithms feasibility,and prove our algorithm superiority.In this thesis,the simulation platform was built using the network simulation tools Mininet and floodlight controller,and a custom network topology was generated,and the load balancing architecture would be built.Through testing the system and sorting experimental data,we found that the classification algorithm based on machine learning can better achieve the load balancing of the SDN network.At the same time,the effectiveness and feasibility of the load balancing routing strategy proposed in this paper are verified.