Client-centric data streaming on smartphones: An energy perspective

Todays users extensively download video files on their wireless handheld devices, namely, smartphones and tablet computers, which are inherently power constrained. The batteries on those devices barely last for 2-3 hours while downloading and playing video files. Experiments have shown that video downloads account for a large portion of the total energy cost of downloading and playing video files. In this paper, we present a novel, energy-efficient, purely client-centric video downloading algorithm with three tunable parameters: buffer size, low water mark, and socket-reading size. By means of implementation of the algorithm on a smartphone and measurements of the actual energy cost of downloading video files, we show the impacts of the three parameters on the energy cost of video downloads. By tuning the buffer size, low water mark, and socket-reading, we observed energy savings of 60%, 64%, and 63%, respectively. Armed with the insights into the process of video downloading, mobile app designers will be better positioned to fine tune their apps to reduce the energy cost of downloading large files in general.