Ultra-low Latency Video Delivery using WebRTC Data Channels

Delivery of video with camera-to-screen latencies of less than two seconds, so called ultra-low latency, is the latest battleground for live streaming providers and vendors. Historically, broadcast latencies have been between five and ten seconds, significantly lower than those which can be achieved using the standardised streaming protocols based on HTTP. The frustration of hearing about a change in score, through social media for example, before seeing it on a streaming service prompted the development of the low latency variants of HTTP based protocols. These can bring the camera-to-screen latency down to between six and eight seconds, easily competing with, if not surpassing, standard broadcast. — Providers’ need for new monetization strategies and increased consumer engagement drives the need to reduce this further. Existing technologies’ inherent limitations in this respect demand a new delivery mechanism. The development of WebRTC Media Streams focuses on video conferencing where round-trip latencies in the 100s of milliseconds are key. However, video conferencing has little need for the high picture quality, content protection, alternative audio, and delivery to millions of users necessary for premium streaming service providers. — This paper discusses the limitations and advantages of using HTTP and WebRTC for media delivery and presents a novel approach to delivering media to clients using WebRTC Data Channels. By using a UDP-based protocol, WebRTC overcomes the congestion and head-of-line blocking issues associated with TCP, allowing the delivery of media with latencies below two seconds. The use of Data Channels has three main advantages over Media Channels. Firstly, it enables video content to be encrypted using standard Digital Rights Management methods. Secondly, it allows the use of broadcast quality encoders which make use of all codec tools rather than the restricted toolset dictated by WebRTC encoders. Thirdly, the use of Data Channels removes the need for a per-client encode of the content. Instead, the pre-encoded and encrypted content is simply replicated to all clients at the edge of the network. This, in turn, means that the solution scales efficiently to the ever-increasing streaming audience.

Published

2023-10

Content type

Original Research

Keywords

Live streaming, WebRTC, ultra-low latency video

DOI

10.5594/M002032

ISBN

978-1-61482-964-5