Practical Bandwidth Allocation for Video QoE Fairness

Abstract

Nowadays, the QoE unfairness problem exists under the scenario of multiple clients sharing bottleneck links, as clients just maximize their Quality of Experience (QoE) independently via adaptive bitrate algorithm and congestion control algorithms provide only connection-level fairness. Improving the QoE fairness among clients, the video providers would jointly optimize the QoE of multiple clients. Nevertheless, existing solutions are impractical due to either deployment issues or heavy computation overhead. Therefore, we propose the practical bandwidth allocation (PBA) mechanism for video QoE fairness in this paper. Specifically, PBA formulates QoE by considering the impacts of both network conditions and devices and reconfigures congestion control algorithm according to the piggybacked QoE. In this distributive manner, PBA can rapidly converge to the bandwidth allocation with good QoE fairness. Real-world experiments confirm that PBA improves the QoE fairness and accordingly increases the minimum QoE of video streams by at least 13% and 15% compared with Copa and Cubic, respectively. Moreover, the performance advantage of PBA becomes significant under dynamic wireless networks.