Client-driven network-level QoE fairness for encrypted 'DASH-S'

Abstract

Adaptive video streams, when competing behind a bottleneck link, generate flows that lead to instability, under-utilization, and unfairness. Recent studies suggest there is also a negative impact on users' perceived quality of experience as a consequence. Two general classes of solution exist. Client-side bitrate adaptation algorithms can improve stability and may achieve flow-rate equality. However, operating in isolation, bitrate adaptation has no ability to establish QoE fairness. Conversely, network services have been shown to achieve stability and quality of experience by managing bottleneck resources. However, the widespread use of HTTPS renders these services ineffective. In this paper we show that QoE can only be achieved when both network and client interact. We do so by a constructive argument, and then architect client-Driven Video Delivery (cDVD) in response. Our cDVD implementation provides a client-level API into the network and builds on software-defined principles. cDVD measurements reinforce our argument and raise new opportunities for exploration.