OPR: SDN-based Optimal Path Routing within Transit Autonomous System Networks

In traditional Autonomous System$({\mathcal{A}}{\mathcal{S}})$, resources are provisioned in advance and statically based on statistical analysis of the traffic crossing every and some key parameters such as peaks. However, this approach yields to a significant wastage of resources given the continuous increasing of traffic volumes and its dynamicity. Traffic engineering can benefit of the Software Defined Network (SDN) scheme separating the control from the data plans. We tackle in this paper the traditional distributed routing limitations within a transit ${\mathcal{A}}{\mathcal{S}}$. We propose a new optimized SDN-based routing algorithm to manage incoming data flows requesting predefined throughput and a maximum accepted loss rate (i.e., application requirements) along its path. The routing problem is formulated as an Integer Linear Program (ILP). To resolve this problem, we propose a centralized SDN application named Optimal Path Routing (OPR) based on Gomory Cutting Planes and Branch-and-Bound Algorithms. Based on extensive SD-WAN emulations mainly built over ONOS/Mininet/Quagga platform, the results obtained show that our SDN based optimization solution largely outperforms the traditional shortest path algorithm in terms of packet loss, latency, jitter and throughput satisfaction rate. We show that OPR, compared with the traditional shortest path, is able to increase throughput by more than 40% and to reduce by more than 90% the latency and the packet loss rate.