Research And Implementation Of A High Scalable And Distributed SDN Control Plane

SDN promoted the innovation and rapid evolution of the Internet, a growing number of institutions begin to adopted SDN network technology to deploy network. In SDN network, the controller is responsible for centralized control of the entire network.Controller is the core of the whole network, once the performance or security of the controller can not get effectively guaranteed, then the degrading of network service ability and even the paralysis of the entire network followed. In addition, from the perspective of the network construction, a single controller cannot cope with problems of SDN network across multiple areas. As a result, there are problems of performance, reliability and scalability in the deployment schema based on single controller which is unable to cope with large-scale network.In order to improve the performance, reliability and scalability of the control plane, there is need to adopt the way of deployment that multiple controllers work cooperatively, this thesis proposes a highly scalable SDN control plane architecture based on the idea of distributed, cluster, load balancing.From the perspective of distributed technology, because the controller in the SDN network has the natural characteristics of centralized control and control the global network state information, it is particularly important to guarantee the consistency of network information among the controllers.So, the key part is to solve the problem of good collaboration between the controller for the proposed control plane architecture, this paper designs somes functions based on atomic operations provideed by Zookeeper which include controller member management, group management, incident reporting and data sharing. What is more, load balancer builded by the OpenFlow switch and controller is used to further improve the ability of handling of the whole system and improve the system scalability.Finally the feasibility of the scheme is verified by simulation experiments, this scheme can achive good reliability and scalability on the premise of performance guarantee...