Optimizing the controller placement problem in SDN with uncertain parameters with robust optimization

Abstract

By separating the network management plane from the data plane, Software-Defined Networking (SDN) is a networking architecture that offers a centralized view of the network. But due to the growth of the network, there must be several controllers. The number of them and their location create the issue of the controller placement problem (CPP). In many studies conducted in this field, it has been assumed that network information is fully available. In practice, due to dynamic conditions, the obtained information has uncertainty. In facing this challenge of uncertainty, there are different approaches, but one effective approach has been robust optimization techniques. In robust optimization, the goal is to make a feasible and optimal decision to optimize the objective function in the worst case. Robust methods are also different, and we use the Bertsimas method. In this article, the uncertainty of the traffic rate parameters produced by each switch as well as the capacity or processing power parameter of each controller will be the problem. Finally, we will show the advantage of the robust method in a scenario. In such a way that if there is a prediction for the deviations that happen during the execution of the network, the actual value of the objective function will be equal or even reduced in some cases. © 2023 IEEE.