A QoE Fairness Approach for Adaptive Video Streaming

Abstract

According to Cisco, video traffic accounts for a mjority of all IP traffic globally and the number and type of internet connected devices is growing. DASH (Dynamic Adaptive Streaming over HTTP) has emerged as an effective way to provide these heterogeneous clients a high quality of experience (QoE) when streaming video. In this thesis, we propose FairQ, a client-server DASH adaptation algorithm based on a systematic analysis of DASH techniques and algorithms. FairQ utilizes client's request qualities, buffer levels, resolutions, and SSIM values to not only provide QoE to individual streaming sessions but to assure fairness across streaming sessions. We compared FairQ to a variety of previously proposed DASH algorithms under a variety of network scenarios including a real bandwidth trace. We found that FairQ was able to achieve better fairness than comparable algorithms with an average increase of 53% while producing comparable buffer levels and quality switches under heterogeneous client scenarios.