Authored by Damien Lucas, Co-Founder and CTO
In a recent article, we consider how the promise of a single file format for OTT distribution could drive content owners and distributors to adopt CMAF (Common Media Application Format) to potentially replace the variety of existing formats they currently need to support if they want to address all user devices. But another important aspect of CMAF could well turn out to be a true game changer. Indeed, CMAF holds the promise of low latency content distribution.
Low latency: what’s the story?
Think a live TV show you’re watching at home is really live? Think again. Throughout the digital TV distribution workflow, small delays add up to create what is called latency. In other words, what happens in the studio is only displayed on your home screen a few seconds later. In digital terrestrial TV, latency is around 5 seconds. With IPTV, latency can be anywhere between 5 and 10 seconds. As for OTT, latency may reach 30 seconds or more. While this is rarely an issue when watching a movie, watching a live sports event can turn into a nightmare, when neighbours start cheering across the building or the score is commented on Twitter before you even get to see the action.
The reason for this is that OTT service providers are streaming media files over third-party IP networks, to their viewers’ homes or mobile devices, with no control over these networks’ bandwidth or current workload. To avoid playback incidents in case of network congestion, the playback software buffers the media stream on client devices. So if the stream is interrupted or slowed down, there are a few seconds of playback stored locally ahead of time to fill in the gap. However, 10 seconds of buffer time also means 10 seconds of latency. So the challenge is to strike the right balance between streaming quality and time offset.
Thinking in chunks
When files are streamed using either the HLS or MPEG DASH video container formats, the video is broken down into segments of 2, 6 or 10 seconds. The length is determined by the content distributor, depending on the network quality expected downstream. During live broadcasts, these segments are streamed one by one, as they are produced from the original video feed. Segments can only be downloaded once they are complete and there is always at least one segment buffered client side. This means that in the case of 10-second segments, viewers may have between 11 and 20 seconds of latency at best. This is a ‘one size fits all’ setup, where all users get the same buffer size, regardless of the network they connect through. In other words, a viewer sitting at home with a TV and set-top box connected to a fibre network will endure the same latency as a user walking down the street watching the same program on a smartphone over a 3G connection.
CMAF: small is beautiful
The CMAF format associated with the MPEG DASH video container makes it possible to break the segment barrier by accessing smaller fragments within a single segment. In this case, buffers created by the playback software on the viewer’s device can be as small as 100 to 200 milliseconds, as opposed to several seconds in the best-case scenario previously.
What the CMAF and MPEG DASH combination also introduces is the possibility for the playback software on a user’s device to automatically select the optimal fragment granularity according to the network’s bandwidth and reliability. In other words, while a user watching a programme on a handheld device over 3G or a low-quality WiFi network will still need a larger buffer and will encounter some latency, another user watching from home on a 4K Apple TV over a fibre network will be able to enjoy near-live viewing while still taking advantage of smooth video playback.
Looking forward, as networks keep gaining quality, bandwidth and reliability, low latency will become accessible to a larger proportion of users, on both connected TVs and set-top boxes, as well as on second-screen devices in and out of the home. Certainly a very good incentive for content distribution networks to go the CMAF way.