Survivability-Based Control Network Backup Methods In Multi-Controller SDN

Software Defined Networking(SDN)improves the functions of traditional switches,decoupling control functions and data forwarding functions,and forming control planes and data planes.The control plane consists of one or more controllers,and the data plane consists of OpenFlow switches.The normal operation of the SDN network relies on normal communication between the controller and the switch.The network that carries the communication between the controller and the switch is called a control network,which is composed of a controller,a control path,and an OpenFlow switch.Wherein,a control path refers to a path composed of intermediate nodes and links through which the controller communicates with a single switch.With the increase of network users an d the expansion of network scale,the requests in the network are more and more dense,the processing capability of a single controller is limited,and it is difficult for the controller to respond to all requests within the limited time.Therefore,a multi-controller SDN network is proposed.In a multi-controller SDN network,multiple controllers cooperate with each controller to manage the whole network.Each controller processes a portion of the network request to meet the processing requirements of a large-scale network.Although the performance bottleneck of single controllers has been alleviated,the survivability of SDN networks still faces severe challenges.Once the components in the control network are faulty,the connection between the controllers and the switches may be invalid.The switches can no longer obtain the forwarding flow entry fr-om the controller,which may result in timeout or even loss of data packets in the network.The faults in the control network are mainly divided into controller faults and control path faults,in which the controller faults will cause all the switches that communicate with them to lose connectivity,and the consequences are very serious.The scenario of control path failure is more diverse.Since the control path is a communication path based on physical topology,multiple control paths may pass through a common node or link.These common node or link failures will result in multiple failures of the control path,which in turn causes one or more switches in the network to lose connectivity with the controllers.For a variety of fault in SDN network,many reasonable researches have shown that a reasonable and effective backup scheme is an important prerequisite for fast fault recovery in the network.Therefore,to improve the survivability of the control network in various network failures,this paper focuses on the backup mechanism of the SDN control network,including controller backup and control path backup.The main research contents are as follows:(1)A hybrid controller backup method for multi-controller SDN is proposed.Firstly,the method proposes a strategy to classify the controllers,reasonably identifying important controllers in the network.Then the method configures a dedicated backup controller for these important controllers,and develops a backup scheme based on the switch migration strategy for other controllers,so that the control network survivability after backup is improved while maximizing savings on backup resources.The research contents of this part mainly include:an evaluation method for controlling network survivability for random link failures;designing a survivability-based backup controller placement method;designing a backup method based on switch migration that can effectively save backup resources.Finally,the approach adopts a heuristic algorithm to improve the efficiency of the calculation.(2)A control path backup method based on shared backup bandwidth in multi-controller SDN is proposed.The method firstly models the failure probability of the link and proposes a control path survivability metric based on the link failure probability.Then the method considers the complete recovery of single link failure and proposes a path backup strategy based on backup bandwidth sharing in face of single link failure.Finally,the method combines control path survivability,control network delay and backup bandwidth overhead to propose a comprehensive evaluation method to evaluate the performance of the backup control path and then configure a reasonable backup path for each primary control path.