In June, the Motion Picture Experts Group (MPEG) announced it had reached the “the first milestone for the Common Media Application Format (CMAF).”
CMAF is a new media streaming format whose standardisation is being led within MPEG by Microsoft and Apple, with a view to making it both cheaper and easier to deliver OTT video streams to multiple devices.
Göran Appelquist, CTO of Edgeware, which builds hardware and software systems for delivery of TV content over IP network, and supports the new standard, believes that progress on CMAF means that “from next year, we shall start to see major services running on more or less a common format. Which I think is a big step because it will make life easier for everyone involved in OTT streaming.”
To understand why CMAF is so significant, you have to unpack how video is currently delivered. Appelquist explains that while almost all streamed video today is compressed using standardised encoding techniques – “it’s either H.264 or H.265 [also known as HEVC] for the video and mainly AAC compression for the audio” – the resulting files have to be wrapped up in an additional layer of information known as a ‘container’.
Containers add timing information to these files so that they can be synchronised at their destination, as well as some extra meta-data – i.e. information about what the files contain. And these containers are not standardised.
Thus, for example, Apple’s HTTP Live Streaming (HLS) protocol wraps or encapsulates the data into an MPEG-2 transport stream (the ‘TS’ container format); whereas MPEG-DASH uses MPEG-4 containers instead (the ‘ISOBMFF’ container format).
Edgeware regards these as the two dominant formats today, and such differences mean that “even though the actual media that you eventually will play back is in the same format,” says Appelquist, “because of these different encapsulation formats you still have to create different versions.”
OTT service-providers either have to do this before the video is streamed (in which case it requires a lot of storage, since you have to create and store all the different versions in case they are needed), or create the different versions on-the-fly as required – which in turn requires extra processing power to re-package and encapsulate the media at the time of delivery.
Either way, concludes Applequist, “there is an added cost associated with this, and also an added complexity to build and operate these systems, since you will have a lot of different flavours that you need to work with. So it’s a pretty complex picture right now if you want to address a broad range of end-users [with OTT services].”
CMAF addresses this issue through the creation of, in effect, a standardised container or encapsulation format. “What we’re hoping for with this new standard is that, of course, it will be a much more homogeneous environment that you need to address, and will simplify things a lot for setting up these kinds of services.”
There remain two flies in the ointment, however, which relate to the fact that, in reality, there exist two additional ‘wrappers’ that still need to be added to a common CMAF container in an OTT environment.
The first of these is the so-called ‘presentation’ layer. In today’s adaptive bit-rate environment, OTT video streams are divided into different chunks with different quality-levels, so that the stream can degrade gracefully from a high bit-rate to a lower one when traffic congestion occurs, rather than just freezing. The presentation layer lists these chunks and their qualities, and communicates with the viewer’s client to allow it to choose which ones it needs at what time, dependent on the bandwidth available.
This layer was left out of the CMAF standardisation process, says Applequist, thus “with the [MPEG-] DASH player you use the DASH way of presenting those segments, and with an HLS player you use the HLS way.”
However, CMAF is interoperable with whichever type of presentation layer is used, and Applequist comments that “the presentation layer is the much more lightweight part of the process, so [CMAF has] solved the big problem – but, of course, it would be even better if you could rely on a common presentation format as well.”
The second fly in the ointment is encryption, which constitutes a much bigger issue, concedes Applequist. “Even though the encapsulation is now agreed upon, there are still two flavours of the encryption, for DASH and for HLS. That must be resolved before it turns into a full standard. Otherwise I don’t think this will be successful.” CMAF is interoperable with different encryption systems, just as it is with different types of presentation layer.
However, the consequences for OTT service providers of having to support two different encryption standards mirror those that flow from dealing with two different containers: you need to create two different versions, because encryption creates completely different sets of files. These two versions of the video stream either need to be created ahead of time – which has implications for storage costs – or encrypted on the fly – which has implications for processing-costs.
“You need to align the encryption as well. That is a known issue, and is being worked on by the standardisation group,” comments Applequist. However, he believes the OTT sector “will definitely have to live with [HLS’s proprietary features] for quite some time.”
One of the intriguing trends to emerge from the CMAF standardisation effort is that the two main video streaming standards now in contention – at least in terms of offering different presentation layers and encryption-modes – pitch MPEG-DASH against Apple, rather than Apple against Microsoft (or indeed Adobe).
Applequist comments that Microsoft has “basically given up on their proprietary [Smooth Streaming] format in favour of DASH”, while “we see very little demand from customers” for Adobe’s HDS streaming format.
That means that “DASH is finally taking off,” observes Applequist, commenting that until now it had been “pretty slow” to gain momentum. “Now we see that it is actually happening. We have a lot of customers who have built services using Microsoft Smooth Streaming as a way to reach browsers, [but] if you want to reach browsers today, then DASH is the way to go and more and more customers are starting to see that and transfer their services to DASH. […] With this CMAF step I think we’re getting closer and closer to having this one format reaching all clients, that has been the wish for quite a long time. But now we eventually see that it is actually happening.”
