Load Balancing Model against Multiple Controller Failures in Software Defined Networks

Abstract

This paper proposes a preventive priority setting model to handle load balancing against multiple controller failures in software defined networks. For each switch, a set of controllers can control it, where only one master controller controls the switch and others are slave controllers. We introduce a priority for each controller to become the master controller. At any time, the controller which does not fail and has the highest priority works as the master controller. Once the existing master controller fails, the assignment of new master controller is automatically obtained according to the priority setting to promptly recover the control. The priority set is decided at the network operation start time to minimize the maximum utilization ratio among controllers in the worst-case of failure patterns. We formulate the proposed model in two different forms, which are an integer linear programming formulation and a min-max formulation. We prove that the preventive priority setting problem is NP-hard. A heuristic algorithm is introduced based on the min-max formulation. The numerical results reveal that the proposed model obtains the maximum utilization ratio comparable to those obtained by two baselines; it provides a faster recovery compared to one baseline and reduces the instability of network operation compared to the other.