Performance Bound Analysis for Crowdsourced Mobile Video Streaming
This work addresses performance benchmarking for crowdsourced video streaming, which is incremental as it builds on existing ABR methods by adding a cooperative framework and bound analysis.
The authors tackled the problem of analyzing performance bounds for crowdsourced multi-user adaptive bitrate video streaming over wireless networks by proposing a novel framework that enables cooperative streaming among nearby users, and they derived effective upper and lower bounds on social welfare performance using a virtual time-slotted system and linear programming.
Adaptive bitrate (ABR) streaming enables video users to adapt the playing bitrate to the real-time network conditions to achieve the desirable quality of experience (QoE). In this work, we propose a novel crowdsourced streaming framework for multi-user ABR video streaming over wireless networks. This framework enables the nearby mobile video users to crowdsource their radio links and resources for cooperative video streaming. We focus on analyzing the social welfare performance bound of the proposed crowdsourced streaming system. Directly solving this bound is challenging due to the asynchronous operations of users. To this end, we introduce a virtual time-slotted system with the synchronized operations, and formulate the associated social welfare optimization problem as a linear programming. We show that the optimal social welfare performance of the virtual system provides effective upper-bound and lower-bound for the optimal performance (bound) of the original asynchronous system, hence characterizes the feasible performance region of the proposed crowdsourced streaming system. The performance bounds derived in this work can serve as a benchmark for the future online algorithm design and incentive mechanism design.