QoE Control for Dynamic Adaptive Video Streaming Over HTTP at Access Point
Yukun Yuan, Shan Lin, Gang Zhou · 2019
Recent research shows that when multiple dynamic adaptive streaming over HTTP (DASH) clients stream Internet videos simultaneously via a wireless access point (AP), they may experience poor quality of experience (QoE). This is because DASH clients in a wireless local area network (WLAN) independently estimate current available bandwidth and make bitrate selections. Lacking coordination in the WLAN causes competition for downlink bandwidth, introducing unnecessary bandwidth fluctuations especially when some clients start streaming. Such bandwidth fluctuations lead to cascading video bitrate changes on each stream, resulting in dramatic video quality variations, in the worst case, for example, video stalls. To address this problem, it is essential to mitigate the bandwidth competition by coordinating DASH clients in the WLAN. In this paper, we propose a QoE control framework at the AP to dynamically allocate bandwidth for each DASH client based on their real-time streaming performance feedback. To deal with uncertainty in Internet traffic besides streaming traffic, we design robust control algorithms to minimize stalls while balancing bandwidth sharing among clients. We implement and evaluate our solution with real systems under various settings. Extensive evaluation results show that our design significantly reduces average video stalling duration by 57.1% and improves the average video bitrate by 42.0% for all clients when compared to state-of-the-art solutions.