Research On Reliability Of Software-Defined Networking Control Plane

By separating the control plane of the network from the forwarding plane,the softwaredefined networking can obtain unified centralized control logic,improve the network management efficiency,enhance the network programmable ability,and enable the network behavior to make more timely and flexible response according to the business needs and network state.As the core part of the software-defined networking,the control plane is mainly responsible for two tasks: one is to translate the application plane request into the specific forwarding behavior of the data plane;the other is to provide the application plane with the underlying network topology and state information.In order to ensure timely and reliable services for network applications in SDN,the control plane must maintain good performance and maintain reliable communication with the switches to ensure that the switch can timely and reliably obtain control information and forward data packets.However,in the actual operation of SDN,a variety of unreliable factors affect the reliability of the control plane,making the switches unable to obtain control information in a timely and reliable manner and affecting the performance of the entire SDN.This paper relies on the National High-tech R&D Program(863 Program)"Software Defined Network Architecture and Key Technologies",starts from the analysis of the impact of different fault scenarios on the performance of SDN networks,with the goal of improving the reliability of the control plane,from the perspective of optimizing controller deployment and switch migration strategy,to carry out the research on reliability of software-defined networking control plane.The main research work and contributions are as follows:1.Aiming at the delay optimization problem of re-establishing communication between the switch and the control plane after the control path is faulty,a fault-tolerant deployment strategy based on delay constraint is proposed to reduce the communication delay between the switch and the control plane after the control path is faulty.Firstly,this paper uses the intra-domain control path backup method to respond to the control path failure.Secondly,considering that the backup control path can not meet the minimum delay constraint,a controller fault-tolerant deployment strategy is proposed,and greedy is designed according to the statistical characteristics of complex networks.The algorithm selects one of the deployment combinations under which the constraints of the controller performance,the initial delay,and the load balancing are meet,and the link failure has the lowest impact on the performance of the control plane,to make the optimization of the communication time delay between controller and switches after link failure recovery.Simulation results show that the algorithm reduces the unreliability of switch nodes by 31% on average,and reduces the maximum communication delay between switch and controller by about 7%.2.Aiming at the problem of how the disconnected switches re-establish communication with the control plane after the controller fails,a triggered switch migration strategy for controller failure is proposed to improve the fault recovery speed and reduce the response time of the control plane to the flow request after the controller fault recovery.This paper uses the static information of the network and uses the queuing theory to reasonably convert the controller load index,controller and switch communication delay index and packet loss rate index into a single delay indicator,and design the genetic algorithm to designate the backup controller for the lost switch in advance to improve the fault recovery speed and optimize the flow response time after fault recovery.Simulation results show that compared with the nearest migration strategy and the migration strategy with the lowest utilization rate,this method can reduce the flow response time by 21%~40% when the overall network load is higher.3.Aiming at the problem of how to move out the switch in the overload domain after the controller is overloaded,a multi-domain switch migration strategy for controller overload is proposed.Considering the shortcomings of load balancing and delay optimization of the existing switch migration scheme,this paper proposes a multi-domain migration mechanism,by considering the current network state and historical statistics,and selecting the optimal migration path and the switch to be migrated to achieve better optimization of load balancing between controllers and communication delay between switches and controllers during the same migration time.The simulation results show that compared with the existing strategy,the average flow request response time of the migration strategy is reduced by at least 12% and the maximum flow request response time is reduced by at least 11.7% during the peak load period.