Research On Key Technology And Application Of Control Plane And Software-defined Networking

Software-defined networking(SDN)is an approach to separate the data plane and the control plane,which is believed to improve the flexibility of the network.Therefore,it has drawn great attentions and become an important part of future networks.Supported by the Natural Science Foundation of China,this thesis mainly investigates the control plane in SDN,with the emphases on load balancing,energy saving,information synchronization and the resources allocation in software-defined satellite-terrestrial networks.Accordingly,the distinct contributions of this thesis can be summarized as follows.1.In order to solve the problems of the design of multiple controllers ar-chitecture and passive load balancing,we take a joint consideration of features in horizontal multi-controller architecture and vertical multi-controller archi-tecture.And we propose a hybrid multi-controller architecture.Based on this architecture,we design a double threshold approach to evenly allocate the loads.Simulation results reveal that the proposed approach can significantly relieve the working load on the super controller and reduce the load jitter of multiple controllers.Moreover,in traditional load balancing method,the load balanc-ing approach is implemented after the over-load problem happens,which intro-duces more time delay.Therefore,we implement the ability of automatic man-agement among multiple controllers,and propose a load balancing approach based on reinforcement learning.Simulation results show the effectiveness of our proposed scheme in the aspects of load variation and time efficiency.2.In order to solve the energy saving problem in the control plane,we propose an M-N policy multiple controllers sleeping mode by switching off redundant controllers when the system is in the light traffic condition.Simula-tion results reveal that our proposed policy achieves superior energy efficiency but introduces extra time delay.Additionally,in order to relieve the extra time delay,we propose a energy saving approach based on topology analysis.By analyzing the topology data,the influence of each controller is excavated.Ac-cording to the analysis results,the system opts to switch off controllers with smaller influence.Simulation results show that the improved method achieves superior time delay,compared with the previous approach.3.In order to solve the synchronization problem among multiple con-trollers in feasibility,security,and scalability,we propose a permissioned blockchain based synchronization approach in the control plane.The permissioned blockchain works as a trusted third party to collect and synchronize network-wide views between controllers.In addition,considering different requirements in the con-trol plane,it is unreasonable to use only one consensus protocol as a best fit one in the permissioned blockchain.We formulate and optimize three consensus protocols to improve the performance,so as to meet the different requirements of the control plane.4.In order to solve the resources allocation problem in satellite-terrestrial networks.We propose a software-defined satellite-terrestrial network architec-ture,which realizes the flexible management of multi-dimensional resources.Then,using network-wide views in the control plane,we classify resources as communicating resources,caching resources,and computing resources.We also formulate the joint resources allocation problem as a Markov decision pro-cess,and use a deep Q-learning approach to solve it.Simulation results reveal the proposed approach achieves superior resources utility.Finally,the investigations of the whole thesis are summarized.Meanwhile,several valuable research interests on SDN are discussed.