Proactive Energy-Aware Adaptive Video Streaming on Mobile Devices

Energy-aware app adaptation enables mobile apps to dynamically adjust data fidelity such as streaming video quality to meet a user-specified goal for battery duration. Traditional energy-aware app adaptation is reactive in nature where the operating system monitors the app energy drain and signals the app to adapt upon detecting energy drain deviation from the pre-specified energy budget which can cause high oscillation and poor quality-of-experience (QoE).

In this paper, we observe that modern power-hungry apps such as video streaming and o oading-based apps already come with sophisticated app adaptation to deal with resource changes such as network dynamics and propose proactive energy-aware adaptation where the user-specified energy budget is integrated with the app adaptation logic. The potential benefit of such an approach is that app energy drain adaptation is no longer an "after-effect", and hence the approach is likely to reduce the oscillation in app adaptation and improve the app QoE.

In this paper, we study the design, implementation and performance tradeoffs of reactive and proactive energy-aware app adaptation in the context of one of the most powerhungry classes of mobile apps, ABR-based video streaming.

Our study shows that proactive energy-aware ABR video streaming is easy to implement by leveraging the built-in adaptation of modern apps and can improve the QoE of reactive approach by 44.8% and 19.2% in streaming 360 o videos to Pixel 2 and Moto Z3 phones under low power budget.