The Load Balancing Research Based On BP Neural Network Algorithm In SDN

In recent years,with the rapid development of mobile Internet and e-commerce,big data and cloud computing become the new trend.The network traffic pressure is getting increasingly higher,and at the same time,the traffic pattern is gradually becoming highly dynamic.The ability to provide bandwidth on demand instead of total transmission capacity gain has become the key to creating profits.In consideration of the construction costs and the diversification of services,the service provider is more eager for the flexible control ability.In addition,IT infrastructure also needs to be changed in a new way to support the rapid and flexible business operations.SDN(Software Defined Network)is brought into being under these circumstances.It is a new type of network architecture,which is a great innovation and breakthrough of the traditional TCP / IP network architecture.As the name suggests,its idea is to achieve the control and management of the network by means of software definition and driver.The architecture is divided into application layer,control layer and data layer.The network equipment focuses on the data forwarding so as to improve the speed.And the control section is under centralized management and scheduling of the SDN controller through programming,which can determine the forwarding and scheduling strategy according to the needs of design program.Developers on the application layer can get the global view of the network through the SDN controller,they can deploy the business flexible and free?However it leads to another problem,that is,how to realize the reasonable allocation of resources while providing quality services,achieve load balancing in SDN.Load balancing is becoming increasingly popular in recent years.In the face of more new applications and more network traffic,the network is also under more pressure.In order to guarantee the quality of service,on the one hand,the data processing ability and response speed should be improved;on the other hand,it is necessary to carry out load balancing to realize the rational allocation of network resources.Therefore,how to use load balancing to improve the performance of the network in SDN has become a new research hotspot.According to the status that the current network suffers from all kinds of pressure and the bottleneck of traditional network architecture,this paper analyses the new network architecture SDN and comes up with a load balancing scheme in SDN based on BP Neural Network algorithm.One of the features of SDN network is the separation of control layer and data layer,the controller can access the whole topology of the network,and can obtain real-time network state information,including link load,delay and so on,we can use this point in the control layer for a routing strategy for packet forwarding.,and to achieve load balancing.We can use heuristic algorithm to obtain a scenario of an approximate optimal solution,but the time required for the cost is very high,so we take it as a training tool,an intermediate product.BP Neural Network algorithm is the most typical and the most widely used algorithm of artificial networks.It has the advantages of self-learning ability and generalization ability,which can carry out model training on input and output relationship,predicting the output based on the input very quickly after training.SO in this paper,we use the real-time network status and Qo S request as the input of BP algorithm,and the the optimal routing heuristic algorithm was calculated as the output of the BP algorithm,and we use BP algorithm to learn and train the model,then we put the model(routing decision)in the SDN controller,so as to achieve better results in load balancing in SDN.