Robust Controller Placement Based on Load Balancing in Software Defined Networks

To further improve Software Defined networks performance, robustness and load balancing, it is valuable to determine how to optimally deploy controllers against links failure. In this paper, we model the robust controller placement (RCP) optimization problem with an integer linear programming, termed RCP_ILP, which optimally places the least controllers to meet the robustness and load balancing. The novelty of our model is taking into controller coverage probability, network transmission efficiency and Gini coefficient of controller loading at the cost of least controllers. To reduce the computational complexity for the optimal configuration, we propose a heuristic RCP algorithm. The extensive simulations conducted with real network topologies show that the heuristic RCP algorithm improves the robustness and load balancing of SDNs against links failure.

I. Introduction

Software Defined Networks (SDNs) are envisioned as a promising paradigm that decouples the control plane and data plane. The control plane is formed by a set of controllers and each of them manages one or more simplified packet-forwarding switches. The separation simplifies the networking management and improves its scalability [1]. Scalability and robustness should be more carefully considered in SDNs. Since network failures could cause disconnections between the control and data planes, and further disable some of the switches, it is of great importance to improve the robustness of SDNs [2].