Performance evaluation of OpenFlow-based software-defined networks based on queueing model.

OpenFlow is one of the most famous protocols for controller-to-switch communications in software-defined networking (SDN), commonly seen as a promising way towards future Internet. Understanding the performance and limitation of OpenFlow-based SDN is a prerequisite of its deployments. To achieve this aim, this paper proposes a novel analytical performance model of OpenFlow networks based on queueing theory. After depicting a typical network scenario of OpenFlow deployments, we model the packet forwarding of its OpenFlow switches and the packet-in message processing of its SDN controller respectively as the queueing systems MX/M/1 and M/G/1. Subsequently, we build a queueing model of OpenFlow networks in terms of packet forwarding performance, and solve its closed-form expression of average packet sojourn time and the corresponding probability density function. Finally, the numerical analysis is carried out to evaluate our proposed performance model with different parameter values. Furthermore, our controller model is contrasted with the classical one by utilizing the popular benchmark Cbench. Experimental results indicate that our controller model provides a more accurate approximation of SDN controller performance.