Modeling and optimization of packet forwarding performance in software-defined WAN

摘要

As a novel network paradigm, Software-Defined Networking (SDN) offers numerous benefits for wide-area networks (WAN), like promoting application performance and reducing deployment costs. However, it also comes along with an inherent penalty to essential network performance such as packet forwarding delay, primarily due to the involvement of logically centralized controllers. This paper is motivated to provide an accurate queueing system of packet forwarding performance in software-defined WAN based on modeling its controller cluster and OpenFlow switches. In particular, we approximate the packet-in message processing of the controller cluster as an M/M queue based on the derivation of its message arrival process. Meanwhile, we characterize the packet processing of an OpenFlow switch as an M/C/1 queue after taking an insight into its packet switching process. As a further step, we build an optimization model of controller cluster deployments to obtain the optimal number of controllers in the cluster. Finally, our proposed queueing model of SDN controller cluster is evaluated with the prevalent benchmark OFsuite_Performance by experiments, and their results indicate that our proposed model provides a more accurate approximation of controller cluster performance. Furthermore, we perform numerical analysis on packet forwarding delay and solve the optimal number of controllers for different varying parameters, which offer effective guidelines for software-defined WAN deployments.