The Challenges of Deploying Apple’s LL HLS in Real Life
About four years ago an Australian telco acquired the rights to the English Premier League. This was the first time that Australia was to experience the distribution of a major league game on a pure OTT only platform. During the launch and for a long time after it, the telco experienced numerous tech problems and complaints about its service and the viewers’ experiences. Most of the reported issues came from an audience who were raised watching sport on traditional TV and who expected the same high quality of service from a telco who had just entered the media market Down Under.
One of the more interesting cases brought forward during this time was a complaint, reported as:
“A disgruntled English Premier League (EPL) fan has raised concerns with the Australian Competition and Consumer Commission (ACCC) over the telco’s claims that games are streamed live.” (Source: https://www.adnews.com.au/news/epl-fan-takes-optus-to-accc-over-live-claim)
The issue here was that the latency between the stadium in England and someone watching the match in Australia was, on average, about 60 seconds. It showed that latency in live sports OTT distribution matters, to quote:
“My friends celebrate the goal that’s just been scored and I won’t see it on my screen for a whole minute…“. (Source: https://www.adnews.com.au/news/epl-fan-takes-optus-to-accc-over-live-claim)
Fast forward to 2019 and not much has changed. Most technologies used in live OTT streaming have not moved on; the focus in video technology has been on improving bandwidth and encoding quality, not on addressing the real fan issue – latency. The reason that OTT companies pick quality as their focus point is largely because addressing latency requires changing many of the parameters outside of their domain. In other words, they cannot influence enough of the distribution chain in order to deliver lower latency at scale. They simply do not own enough of the typical video production, encoder, packager, CDN, player or operating system chain to progress to something which really provides value for sports fans.
In reality, there are about three companies in the world who are able to influence the OTT industry to bring about the changes needed to really make the Aussies happy, and every other sports fan in the world today and in the future.
Earlier this year M2A Media joined Apple and a small number of other companies to start working on all the steps in the distribution chain to bring latency down to the same level as traditional broadcast and even beyond. There have been many attempts in the past but none that were supported by a giant like Apple, who is able like no other, to push the industry forward for the benefit of everyone. To enable the necessary change in the chain Apple has extended the HLS standard used in OTT live streaming with a number of new things.
1) Segments can now be sliced in smaller Parts.
2) Parts do not have to start with an IDR frame
3) HTTP2 Push is used to distribute video content to clients
4) Playlists can be ‘delta-ed’
5) Playlist responses can be held by the server until a future point in time
In other words, Apple has changed the encoding and packaging rules slightly, but their biggest change is the requirement to use HTTP/2 for low latency. This protocol has been around for a number of years but Apple is one of the few companies in the world that owns a billion client eco-systems (iPhone, iPad, Apple TV) and that can pull the network and CDN video industry together and demand the use of HTTP/2 for video distribution for things that really matter to sports fans. Both the protocol and the requirements formulated by Apple do result in more satisfied fans, and this helps all of us in the industry.
Adopting HTTP/2 is a big deal; the push property used in the protocol goes against anything that has been commonly used in the OTT industry so far. So why is it so important and why does it make things work so much better?
Diagram A: Standard OTT Distribution
In a standard live OTT flow ad playout source is connected to an encoder which chunks the live video stream in six second segments of different resolutions and quality. Each of these segments are distributed to clients based on a client requesting a certain segment starting at time X and at quality Y. These requests are calculated on the client based on-screen size, network conditions, costs etc. The important thing here is that segments are announced in segment playlists and these playlists are cached for the duration of a segment across an OTT distribution network; the more steps in the CDN chain, the greater the latency. Adding HTTP/2 Push to the available toolkit addresses this inherent latency problem. The player does not need to request video content, it is being pushed and replaces the request model which reduces latency.
In low latency mode Apple instructs the player to request a Playlist in the future and asks the server to push the video content along the Playlist response when that future point in time has been reached.
Diagram B: Low Latency OTT Distribution
Two important things are visible in the above:
1) The player no longer requests the video content itself, which reduces the network round trip latency.
2) Since the content is pushed from the packaging stage directly to the client, but is still cacheable. The traditional player loop ‘wait for the next segment announcement in the Playlist’ no longer exists.
Adding short duration Parts, typically around 300 ms in duration, to the mix results in an end to end latency of about 2-5 seconds which is a huge improvement compared to what is achievable with the traditional toolbox.
In order to deliver this new low latency workflow at scale M2A Media is working with Media Excel, Apple, DAZN, Unified Streaming, Akamai and Stackpath.
Valentijn’s second blog in our LL HLS series, The Challenges of Deploying Apple’s LL HLS in Real Life: Part Two to follow soon. Stay tuned!
About the author
Valentijn Siebrands began developing live OTT services early on. As Solutions Architect at M2A Media he brings with him 20 years experience working with international broadcasters and telcos.