Abstract:
In order to improve the scalability and reliability of the control plane, multiple controllers are introduced in software-defined network. However, some performances, such as controller capability and network stability, will get poor. In order to solve these problems, a load balancing (Game Theory Load Balancing, GTLB) mechanism based on game theory is proposed to improve the load imbalance of multiple controllers. First, the overloading controller invites some adjacent slave controllers to participate in the game as players and form a game domain. Then, the target controller is determined by the particle swarm optimization algorithm based on the delay between the controller and the switch and the migration cost of the switch. And the migration switch is selected according to the node distance and flow request volume. And the orderly and seamless migration of the switch is realized according to the migration timer. In comparison to the existing balancing mechanism, the proposed mechanism reduces the total communication overhead of the network, shortens the average flow establishment time by 0.12 s, and improves the utilization rate of the controller by 20.4%. The load balance performance of controller is improved in the GTLB.