Consistency, Feasibility, and Optimality of Network Update in SDNs

Abstract

In software defined networking, to maximize the network utilization, its control plane needs to frequently update the data plane via flow migration as the network conditions change dynamically. Since each switch updates its flow table independently and asynchronously, the network state transition may result in serious link congestion and packet loss if it is done directly from the initial to the final stage. Deadlocks among flows and links may also block the update process. In this paper, we pay close attention to elaborately resolving deadlocks with the help of spare paths during the network update. We prove that the feasibility of the consistent flow migration can be determined in exponential time. Furthermore, we demonstrate that even if there are multiple consistent migration plans, finding the optimal one that occupies the least leisure bandwidth resources is NP-hard. For the case in which no consistent plan is found, we introduce an efficient method to rate limit flows in order to reduce the packet loss. Extensive simulations show that our solution achieves a much smaller traffic loss rate at the cost of affordable spare link resource usage compared to prior methods.