RESA: A Real-Time Evaluation System for ABR
Yanan Wang, Haili Wang, Jiaoyang Shang, Hu Tuo · 2019
Adaptive Bitrate Algorithm (ABR) is a category of important technique to improve the Quality of Experience (QoE) of video streaming service. Many ABR algorithms have been proposed recently and achieved good results. However, there are still some problems to be solved: (1) These ABR algorithms are based on open source data and evaluated by only one simulated player; (2) The QoE metrics are defined differently and their parameters are set by experience with little explanation; (3) These algorithms consider only the QoE of users but ignore the bandwidth cost of video service providers. In order to solve these problems, we propose a Real-time Evaluation System for ABR (RESA) which do evaluations by real online users. A reasonable QoE metric is also introduced by setting its parameters based on user preferences during video watching. We evaluate the state-of-the-art ABR algorithms on RESA with real video streaming service. Finally, we further introduce bitrate controlling into the Adaptive Bitrate Algorithm to solve the bandwidth cost problem for online ABR.