Congestion-aware MAC Layer Adaptation to Improve Video Teleconferencing over Wi-Fi
Recently in March, Liangping Ma, Member of Technical Staff of the InterDigital Labs group, demonstrated one of InterDigital’s cutting-edge network optimization technologies at the ACM Multimedia Systems Conference (MMSys), in Portland, Oregon. Ma worked jointly with Wei Chen, Senior Engineer, InterDigital, and Chien-Chung Shen, Professor of Computer and Information Sciences, University of Delaware.
The network optimization technology addressed a major challenge to video teleconferencing on mobile devices that communicate over Wi-Fi. Unlike video streaming of prepared content such as Netflix movies, video teleconferencing requires the network to provide very low latency. In addition, Wi-Fi is prone to packet losses resulting from either channel errors or network congestion, even with the retransmission mechanism in Wi-Fi. The-state-of-the-art technologies are either too slow in providing feedback on packet losses (e.g., RTP-layer retransmission based on end-to-end feedback) or too inefficient in using network resources (e.g., application-layer forward error correction).
The InterDigital Labs group tackled the challenge by looking at how to enhance the Medium Access Control (MAC) layer of Wi-Fi. At a first glance, it may be tempting to keep sending a lost packet until it is successfully received. This approach has two problems:
- When the Wi-Fi network is congested, to keep sending a lost packet will only make the congestion worse.
- To keep transmitting a lost packet will hide congestion-caused packet losses that are essential to higher-layer congestion control protocols such as the Google Congestion Control protocol in WebRTC.
Instead, the InterDigital Labs group proposed to first detect congestion, and then increase the number of allowed retransmissions if and only if congestion is not present. In other words, the number of allowed retransmissions is increased if and only if a packet loss is caused by channel errors. The technology significantly reduces packet losses without disrupting congestion control. The benefit is clearly demonstrated by the much smoother play of the video.