Adaptive QoS for data transfers using software-defined networking

Abstract

The research community has devoted much attention to the use of classic network technologies for advanced traffic engineering, but the use of software-defined networking (SDN) for this purpose, particularly in regards to Quality of Service (QoS) optimizations, remains relatively unexplored. We have developed a QoS framework that leverages SDN capabilities to achieve optimal throughputs for all QoS flows on a congested network. Specifically, we propose an approach that reassigns bandwidth unused by one or more flows to other flows in priority order, while simultaneously ensuring that QoS flows can reclaim their reassigned bandwidth whenever they are able to use that bandwidth. At the heart of our framework is the Adaptive Quality of Service (AQoS) algorithm, inspired by Integrated Services (IntServ) principles, which enables fine-tuned, real-time control over per-flow bandwidth allocations, combined with classic Differentiated Services (DiffServ) priority classes that determine how bandwidth is distributed. We evaluate and compare the results of the AQoS throughput optimizations to those achieved by several baseline algorithms common to both classic and SDN networks, with promising results.