Minimizing Congestion Impairment of Network Update in SDN: A Flow-Based Solution

Abstract

In software defined networks, network states are frequently updated by controllers. Unfortunately, due to resource and time constraints, there are scenarios in which transient congestion and packet loss are inevitable. In this regard, minimizing the packet loss ratio becomes crucial. Previous efforts on congestion-free updates suggest link-based solutions, which aim at minimizing the overloaded data volume on the bottleneck links. Observing the fact that the least overloaded data volume on links still does not guarantee the least packet loss, in this paper, we propose a flow-based update solution that directly minimizes the packet loss by jointly optimizing the congestion duration and rate limitation. Specifically, congestion impairment is defined to jointly accommodate the flow's importance and packet loss. Then, we present the FBU (Flow-Based Update problem), which minimizes the congestion impairment on a flow basis. To deal with the NP-hardness of this optimization problem, we propose MIC, which is an efficient two-phase heuristic algorithm based on the relationship between rate limitation, congestion duration and packet loss. Experimental results show that MIC can reduce up to 84% of packet loss compared to previous algorithms.