Optimal Fairness and Quality in Video Streaming with Multiple Users

Abstract

With the majority of video distribution services relying on the HTTP adaptive streaming paradigm, a great deal of research is geared towards developing algorithms and solutions for improving user perceived quality while making efficient use of available resources. Our goal is to provide the means for benchmarking such solutions in the context of multiple users accessing Video on Demand content while sharing a bottleneck link. For that purpose, we propose a quadratic problem formulation to compute the theoretical optimum in terms of adaptation strategies and corresponding segment downloads across multiple users under given bandwidth constraints. By aiming to maximize both service quality and fairness, we quantify and compare the impact of different fairness objectives (bandwidth fairness, pattern fairness, and session fairness) on resulting quality and achieved QoE fairness. Based on conducted simulations and parameter studies, our results demonstrate the benefits of optimizing for session fairness as compared to other approaches.