Main and Secondary Controller Assignment With Optimal Priority Policy Against Multiple Failures

Abstract

This paper proposes a master and slave controller assignment model against multiple controller failures in software defined networks considering latency between switches and controllers. The survivability guarantee of each switch is satisfied by assigning a set of controllers, where one of them works as the master controller to control the switch. Given assigned controllers, we introduce a policy-based approach to automatically specify the master controllers in each failure pattern, which leads to a lightweight configuration on a switch. We define the average-case expected latency, the worst-case expected latency, and the expected number of switches within a latency bound, as three objectives to be optimized in three different problems. We prove that a low latency first policy achieves the optimal objective for each considered problem. We formulate the proposed controller assignment model with different goals as three mixed integer linear programming problems. We prove the NP-completeness for all the three problems. A greedy algorithm with polynomial time complexity is developed; we show that it provides a 1/2-approximation for the case without the survivability guarantee constraint. The numerical results observe that the proposed model obtains the optimal objective value with the computation time about 102 times shorter than that of a baseline that introduces decision variables to determine the master controllers.