Balancing Quality of Experience and Traffic Volume in Adaptive Bitrate Streaming

Takuto Kimura,Kazuhisa Yamagishi,Arifumi Matsumoto,Tatsuaki Kimura

doi:10.1109/access.2021.3052552

Takuto Kimura, Kazuhisa Yamagishi + Show 2 more

Open Access

https://doi.org/10.1109/access.2021.3052552

Copy DOI

Abstract

Adaptive bitrate (ABR) streaming services have spread with advances in the codec, video streaming, and network technologies. For smooth video playback in ABR streaming services, a video player runs an ABR algorithm, which dynamically adjusts the bitrate of video data on the basis of the statuses of the network and player. Existing ABR algorithms calculate a suitable bitrate to maximize the quality of experience (QoE). However, providing a high-QoE video increases network investment costs and content delivery network (CDN) usage fees. According to a survey conducted by the Streaming Video Alliance, mobile users prefer low-traffic videos to high-QoE videos. To reduce traffic volume, commercial video-streaming services enable users to set an upper limit of the bitrate. However, this cannot always achieve the required QoE because they cannot select a high bitrate even when the communication environment improves during viewing. In this paper, we propose BANQUET, a novel ABR algorithm that can reduce the traffic volume while maintaining QoE above the target QoE. The target QoE can be set by users or streaming providers considering user's preferences or CDN budget. BANQUET selects a suitable bitrate by estimating QoE and traffic volume that will be experienced by all the bitrate patterns for the next several chunks on the basis of future throughput and a buffer transition calculation. The trace-based simulation showed that BANQUET reduces traffic volume 18.3%-51.2% on average in the mobile environment and 1.2%-38.3% in the broadband environment while maintaining QoE the same as or better than existing algorithms.

Highlights

Video traffic volume is expected to grow four-fold from 2017 to 2022 and account for 82% of all traffic in 2022 [1]
Traffic volume is reduced by 18.3%–51.2% in the mobile environment and 1.2%–38.3% in the broadband environment on average, while maintaining the same or better Quality of Experience (QoE)
We evaluated BANQUET through a trace-based simulation using actual throughput data collected in mobile and broadband environments

Summary

INTRODUCTION

Video traffic volume is expected to grow four-fold from 2017 to 2022 and account for 82% of all traffic in 2022 [1]. If the ABR algorithm chooses a higher rate when we use these analysis-based methods, it may achieve a higher QoE than the users or streaming providers need and result in an increase in traffic volume. The setting manners are different, these services reduce the traffic volume by setting the maximum bitrate that ABR algorithms can select These methods may not always achieve the user’s required QoE. BANQUET calculates the suitable bitrate series that minimizes the traffic volume while maintaining QoE higher than the target QoE, Tq by using information from the video player. The details of this step are given in Subsection III-B. We evaluate the calculation time and analyze the decision policy of h and Td to reduce the calculation time in Subsection IV-D

14: Calculate qt

EVALUATION

CONCLUSION AND FUTURE WORK