Research And Implementation Of SDN Distributed Control Mechanism

Software Defined Networking(SDN)has the advantages of easy management and programmability,and can centrally manage the network,but it brings up some issues of scalability.The industry proposes the multiple SDN-controllers architectures.However,most of the architectures can not dynamically adapt to the change of network traffic and causes a load-imbalance among SDN-controllers due to the statically configured mapping between a SDN-controller and a SDN-switch.By analyzing the features of the current multiple SDN-controllers architecture,this thesis proposes a multiple SDN-controllers control mechanism for load balancing,which can dynamically adjust the relationship between a SDN-controller and a SDN-switch.According to the functions of the different OpenFlow messages and the role of the SDN-controller,this thesis designs a switch migration mechanism.This thesis proposes a controller adaptive algorithm,the controllers can be dynamically added and deleted,and switches can be migrated according the change of network traffic.The algorithm takes into account the controller load and the transmission delay between a SDN-switch and a SDN-controller.Meanwhile,it reduces the number of migration as much as possible.This thesis implements the multiple SDN-controllers control mechanism,including load measurement module,decision module and switch migration module.The load measurement module periodically collects and calculates the load of the controller.According to the load of the controller,the network traffic request and the set threshold,the decision module determines what to do.By modifying the role of the controller,the migration module achieves migration of the switch.The results show that the multiple SDN-controllers control mechanism improves the control plane scalability,but also realizes the load balancing between the controllers.The results also show the algorithm can dynamically adjust the number of controllers and the mapping with SDN-switch according to the change of network traffic,and improves the controller resource utilization.Compared with the nearest migration and free resource utilization minimum migration strategy,this algorithm can balance the load of the controllers better.