Incremental SDN Deployment to Achieve Load Balance in ISP Networks

Abstract

Software defined network (SDN) decouples control planes from data planes and integrates them into a logically centralized controller. With capture of the global view, the controller can dynamically and timely reply to the changes of network states. However, replacing the entire traditional networks, e.g., Internet Service Provider (ISP) networks, with SDNs is difficult and computationally expensive. Hence, incremental deployment of partial SDN devices has received much attention. In this paper, we consider the k-LB problem, i.e., upgrading at most k legacy switches to SDN switches to achieve load balance. We claim that k-LB problem is NP-hard and there is no polynomial time $(N+M)^{1-\epsilon }$-approximation algorithm for any constant $\epsilon > 0$ unless $\textbf{P} = \textbf{NP}$, where N (M) is the number of switches (links) in the network. Given these negative results, we propose an effective greedy algorithm and claim that it reaches an approximation guarantee of $\frac{c_{avg}}{c_{min}} M$, where $c_{avg}$ ($c_{min}$) is the average (minimum) link capacity. Large-scale simulations on real ISP networks show that our greedy algorithm achieves near optimal performance and decreases the maximum link utilization by 30% on average compared with the state of the art.