With the OTT Summit in Madrid drawing close, SportsPro spoke to Franck Chevalier, Limelight Networks' director of solutions engineering, to get his take on some of the technical aspects of digital broadcast delivery.
What are the hurdles to cross before digital overtakes linear distribution?
There are both technical and business challenges to overcome as online video streaming growth continues. Current streaming infrastructure cannot keep pace with streaming demand. The issue is unicast distribution – one stream to one device vs broadcast delivery. Over the next few years, streaming capacity will have to grow to two or three times what exists today, with greater 4K adoption only set to make this worse.
Consumer expectations for video quality have been set by broadcast TV. Viewers want a consistent experience across viewing devices and platforms. Rebuffering issues are too common today.
Additionally, the online video landscape is overwhelmed with content, making search a huge issue. Finding something to watch across all the subscriber service ‘walled-gardens’ is too difficult. New search tools are needed that can search across all the various platforms.
And finally, viewers need too many service subscriptions to cover all the content they want to watch. There is a limit to how much they are willing to pay per month.
Where does latency sit on the spectrum of digital streaming issues? Should it be a bigger priority?
Streaming latency is an issue for specific online video use cases – live sports, interactive live trivia, auctions, gambling, etc. The latency requirements for these vary by use case – but considering all the challenges mentioned above, it should not be a high priority. Existing streaming technology can address most of the latency requirements, and new technologies in proof-of-concept testing today will address the most stringent latency requirements.
What are the technological advances happening to deal with that issue? Will web real-time communication (WebRTC) be a game changer?
A common complaint with online live streaming viewing of sports events is the social media ‘spoiler’ text messages from viewers of the traditional TV broadcast of the same event, ruining exciting action that the online viewer will not see for often a minute or more later. Typical broadcast TV latency is in the range of six to ten seconds. HTTP live streaming (HLS) and dynamic adaptive streaming over HTTP (DASH) can match this latency by using small chunks in the two to three second range. We are seeing this implemented today.
To reduce latency further, common media application format (CMAF) supports a low latency mode that uses micro-chunks, and video playback can begin even as video segments are being formed. In production testing, latency in the two second to four second range has been demonstrated. This will satisfy latency requirements for many interactive use cases for live online auctions and gambling.
The most stringent latency requirements come from online live sports with in-event betting. Attendees at a sports event can be wagering on their mobile devices along with remote viewers. To level the field among these viewers, the streams need to be delivered with sub-second latency. This is where WebRTC can play. WebRTC can really deliver latency in sub-seconds, and is demonstrating this in numerous proof of concept tests today. This will be a game changer for interactive use cases where real-time, more social experiences will be enabled.
What are the challenges of dealing with scale?
Online video is distributed as unicast one stream to one device. As streaming growth continues, distribution infrastructure has to grow with it – an expensive proposition. One way to address this is more multicasting of video streams. But, multicast assumes that everyone receiving the stream has the same bandwidth capabilities and that they’re using a similar device to access the content. This makes it difficult to scale and cater to viewers who have lower or fluctuating bandwidth, or different devices than the stream is intended for.
Also, from the stream monetisation viewpoint, the advantage of internet protocol unicast delivery is personalisation of ad insertion. There may be a hybrid approach that could work. If video content distribution networks (CDN) are used, deploying some level of multicasting within the CDN to distribute streams to the global points of presence (PoP) edges would reduce the load. Where streams are distributed to viewers through local Internet service providers are where the personalisations could be handled by edge resources.
What do you see as the best technological innovations that have come from the advent of OTT?
Firstly, content-aware encoding (CAE) is an innovative technology that is currently used by Netflix. With CAE, encoders adjust encoding parameters in real-time based on video complexity. This provides the highest scalability – ie., encoding speed - but not the lowest compression ratio.
Secondly, CMAF, which reduces the amount of different media file formats required. Since it can be used by both DASH and HLS delivery formats with a common encryption scheme (CENC), CMAF greatly simplifies the OTT distribution at scale. CMAF also provides an option for low latency chunk (LLC), enabling the delivery of segments by small chunks. This means the decoding process can start before a complete segment is encoded, packaged, and received, enabling an end-to-end latency of three seconds or less, levelling the playing field with broadcasters for delivering live sports.
Lastly is edge scaling. The concept of scaling at the edge is simple - CDNs deploy edge nodes inside the ISP access network. For live events, predominantly sports, scaling close to the subscriber will provide the greatest quality of experience.
What OTT Summit session are you most looking forward to?
Delivering live sports at scale: Business and technology working together to thrive in the OTT world.
Franck will be speaking at the OTT Summit in Madrid, over 19 - 21 November, click here for more details.