Research On Control Layer Failure Detection And Recovery Algorithm In SDN Framework

Software Defined Network(SDN)has transformed the network into a resource that can be flexibly deployed,by abstracting the control functions of network devices to form a control layer.The SDN architecture effectively reduces the complexity of network management,and provides powerful programmability for the network to meet the different needs of network users for network resources.In the early stage of SDN,only one controller was deployed to control the entire network.At this stage,as the network scale continues to expand,the network environment becomes more and more complex.The entire network relies entirely on one controller,which is difficult to guarantee the reliability of the network.In order to solve the performance bottleneck,single point of failure and scalability of a single controller,the SDN multi-controller architecture came into being.The multi-controller architecture provides a good foundation for failure detection and recovery of the control layer while improving network performance.The main research content of this thesis is failure detection and recovery of the control layer.In the aspect of failure detection of control layer,the existing failure detection algorithms of controllers mostly uses a fixed timeout value.The setting of fixed timeout value is very difficult,and its size directly affects the performance of failure detection algorithms.Therefore,an adaptive failure detection algorithm is designed in this thesis,which comprehensively considers the CPU utilization of the controller and the detection response delay to dynamically adjust the detection delay.Finally,compared the designed algorithm this thesis with the prediction effect of the gray prediction algorithm and the exponential moving average prediction algorithm.The prediction accuracy of the designed algorithm this thesis is significantly higher than other comparison algorithms.In terms of controller failure recovery,it is an effective way to realize control layer failure recovery by migrating switches in failure domain to the target controller.The existing switch migration mode is single,and the selection of the target controller is mostly based on the switch-controller delay or controller load,which easily leads to the rigidity of the selection of the target controller.It is difficult to guarantee the good comprehensive performance of the control layer after a failure recovery.Therefore,A control layer failure recovery algorithm is designed in this thesis that considers the switch-controller delay,controller load and link failure rate to select the target controller.Finally,compared the designed algorithm in this thesis with the Survivor algorithm and the PPF algorithm in switch-controller delay,controller load and link failure rate.The results show that the designed algorithm in this thesis effectively balances the load between the target controllers under the condition of maintaining the link failure rate and the switch-controller delay.The overall performance of the designed algorithm in this thesis is better than other comparison algorithms.