SEWS: A Channel-Aware Stall-Free WiFi Video Streaming Mechanism

The rise of video streaming has placed significant demands on network infrastructure. These demands are most acutely felt in the wireless space where limited resources are available. Compounding the matter, most techniques for adapting to network dynamics have been developed with wired networks in mind thus making performance in congested wireless networks, especially WiFi, quite problematic. In this paper, we propose a novel cross-layer design to improve video bitrate selection by incorporating MAC layer information. We design a lightweight channel characterization method that can provide an accurate airtime estimation based on the observation of WiFi control packets. We then devise a bitrate adaptation algorithm that can judiciously avoid faulty bitrate increases whenever severe channel competition is detected. Through extensive lab experiments, we show that our proposed method can significantly reduce video stall rates by up to 30x (from 65% to 2%) compared to existing methods.