Reserch On Load Balancing Mechanism Of Low Orbit Satellite Network Based On SDN

With the continuous maturity of satellite communication technology,the advantages of satellite network gradually appear,especially the characteristics of rapid response and global coverage in low orbit satellite network,which will occupy a huge advantage in the development of satellite mobile communication in the future.Based on the traditional structure of network,it is difficult to control the huge satellite network flexibly.At the same time,with the increasing number of satellite network users,the load of network is increasing,but most of the existing load balancing algorithms are difficult to apply to the complex and changeable satellite network.In this paper,the Software Defined Network(SDN)architecture is introduced.According to its characteristics of separation of control and forwarding and open interfaces between layers,using SDN controller to control the satellite network centrally,and then studying the load balancing problem in the low orbit satellite network by SDN architecture is to alleviate the congestion of network and to improve the quality of network service.The details are as follows:Firstly,the typical low orbit satellite networks are studied,and the structural characteristics of One Web satellite network are analyzed.The defects of traditional satellite network are pointed out,and the concept of software-defined network is introduced.Secondly,the characteristics of software defined network controller are analyzed,and the architecture characteristics and advantages of ONOS controller are emphatically discussed.In order to adapt to the algorithm of path planning and the method of the ability of ratable flow splitting proposed in this paper,some extension modules are designed based on the infrastructure of ONOS controller to improve the functions of the acquisition of network information,path planning and the delivery of flow tables.Then,an improved algorithm of reinforcement learning is proposed to make the controller can gradually learn the strategy of path planning through the feedback of the network environment and can calculate the path which is more suitable for the current network.The simulation experiment proves that the path which is planned by the algorithm of improved reinforcement learning can improve the network performance.Finally,in order to make the effect of load balancing more obvious,on the basis of calculating the transmission path,a method of the ability of ratable flow splitting is proposed from the perspective of controller,which solves the problem of the waste of resources in the flow tables and enables data flow to be sent to multiple paths selectively according to the current path state of the network.The simulation results prove that the method of the ability of ratable flow splitting can alleviate the problem of uneven network load.