Blog Post by Charlie Kraus
Over the past year live streaming of major sporting events has taken off. From playoffs and championships in professional sports, and increasingly regular season games, audiences seem to love it.
Why have sports leagues embraced streaming so enthusiastically? For example, in the NFL, viewership of regular season broadcasts has slipped in the past two years. The big TV networks have locked up mulit-year broadcasting rights for billions of dollars. With TV ratings slipping, they need more viewers to bring in more ad revenue. Voila! Live streaming simulcast with the broadcast is adding millions of additional viewers. Sounds like a great solution, right?
Well, most streaming viewers have experienced a problem with this. You’re watching a game on a mobile device as you walk into a sports bar to meet friends, and just as you go through the door everyone in the bar is cheering a just happened score, or you could have seen social media posts reporting it ahead of it showing up on your screen. Yet you experienced the latency gap between broadcast and streaming. This blog will discuss what’s going on here, and the new technologies coming to bear to solve the latency gap. There’s nothing like having a lot of revenue on the line to encourage innovation!
Typical latencies delivering HLS or DASH over the internet are in the 30 second to one minute range. This is because these are HTTP-based protocols, which stream chunks of data. Because each chunk is generated and viewed in real-time, chunk size is a significant part of latency. For example, the default HLS chunk size is 10 seconds, leading to a delivery latency of up to 45 seconds when CDN ingesting, transcoding, distance between source and viewer, and delivery are factored in.
An obvious approach to lowering latency is to reduce the chunk size. This is exactly what we have done with our Video Acceleration configuration option in the Content Delivery Service, which is designed to accelerate the delivery of very small video chunks, and dynamic manifest files. (See more about video delivery services here). This capability is targeted at organizations delivering HLS and DASH live streams from their own infrastructure that the Limelight CDN uses as origin for cache fill. This will reduce live streaming latency down to as low as 5 seconds, a significant improvement over existing HLS and DASH live streaming delivery solutions. There are several successful in-production deployments of Video Acceleration that satisfy use case latency requirements. An example is an Asian online gaming company using Video Acceleration to improve their users game play experience by reducing the latency perceived in play.
The industry focus on solutions for low-latency streaming is a technology called WebRTC, which is supported by all the popular browsers. Detailed coverage of WebRTC is beyond the scope of this blog, but there are multiple WebRTC-based solutions available that can be integrated with CDN network infrastructure. Limelight is active in evaluating this technology, with the promise of providing even lower latency than our current Video Acceleration. Stay tuned for more information about this exciting development.
Video Acceleration is applicable in these scenarios:
Reduce the latency difference between a TV broadcast and the online stream delivery
Consistent viewing experience across multi-devices for non-simulcast delivery
Scaling the delivery of low-latency live video delivery to large audiences
Ensuring a consistent reduced latency experience to users across a range of devices