����tv

What we're doing

If you have access to a smartphone – and 94% of UK adults do, according to – you have almost certainly consumed media from a mobile phone network, whether that is a television programme on , a film or just a clip of your friend’s cat shared on social media. You may also have participated in a video call with friends or family using one of the popular multimedia calling applications. The oft-quoted statistic is that over seventy per cent of all traffic on mobile networks is now video content, but even this figure is becoming out of date as the proportion gradually edges closer to eighty per cent. The number one gripe from users of video streaming services is drop-outs and buffering events that mar their Quality of Experience. Consumption of audio content – such as one of the excellent podcasts or live radio stations available via our popular app – accounts for a lot less data, but listeners are much more sensitive to any disruption in playback.

Why it matters

You may be lucky enough to have a 5G mobile phone and, as cellular mobile network operators roll out their 5G networks, the magic “5G” icon might sometimes light up at the top of the screen to tell you that you have successfully connected. Governments started issuing spectrum licences for 5G use in 2018 and the roll-out of this new technology started in 2019. The promise of 5G is faster broadband mobile data connections, and this is achieved by a massive re-engineering of the network, which has two main components:

• The Radio Access Network provides a wireless link between your smartphone (User Equipment or simply UE in the jargon) and the cellular mobile base station. In 5G, a allows the radio links to operate at higher frequencies and higher data rates using clever signal modulation techniques. The logical function of a base station in a 5G network is called a gNodeB.

• The Core Network provides all the back-end support functions that turn a set of base stations into a useful network that supports services such as making telephone calls, sending text messages, and connecting to the internet. The core’s Control Plane provides registration services for UEs, billing and allocates “slots” in the wireless links to different users. And the Data Plane is responsible for forwarding data back and forth between UEs on the same or different networks. In the 5G Core, these functions were extensively redesigned to use a service-based architecture that can take advantage of edge computing facilities in distributed data centres to support a new generation of advanced communication services.

So that UEs from different manufacturers can work successfully with the equipment deployed in different cellular mobile networks around the world, they all follow a set of standards set by a group of equipment vendors, network operators and content providers called . They agree technical specifications for interoperation and publish them identically through the standards delivery organisations responsible for different regions of the world. This has been the model for the last thirty years and works to the benefit of all, including you the consumer, allowing you to take your smartphone anywhere in the world and connect to a local network there.

Most 5G network deployments for the first four years involved operators upgrading their base station sites to support 5G New Radio, but the gNodeB was simply plugged into a 4G core. These non-standalone 5G deployments might go faster, but they are ultimately limited by the capabilities of the legacy 4G network to which the base station is connected. Only recently have mobile network operators starting to deploy true standalone 5G networks where the gNodeB is fully integrated into a 5G Core. This is a challenging upgrade, but one that will reap enormous benefits in the future.

Like motorways, data networks have historically tended to fill up with traffic over time. Better road connectivity enables new journeys to be made and, analogously, better network connectivity enables new services that weren’t previously possible. In both cases, network capacity is limited and therefore valuable: there is only so much countryside we are prepared to sacrifice under tarmac, and usable radio spectrum is similarly in short supply too. As more and more traffic piles on to the network with new users joining the party, the available capacity becomes saturated, and the network becomes congested. Throughput drops as more and more users try to share the capacity and network efficiency drops as it gets clogged up. Like a car stuck in a Bank Holiday traffic jam, your data may be severely delayed or (where the road analogy breaks down) discarded altogether in an attempt to get the traffic moving.

So, even if your smartphone has a near-perfect connection to its nearest cellular base station (displaying four bars of signal strength), you may still be getting a terrible Quality of Service (QoS) from the mobile network because you are sharing the capacity with too many other users. This is especially evident in high-density urban environments, such as railway stations during the rush hour. Quickly downloading a film before your aeroplane taxis off the parking stand can be a nail-biting race before the attendant asks everyone to switch their devices into “flight safe” mode. Moreover, cellular mobile networks are increasingly being used to provide broadband internet service to isolated locations in sparsely populated rural areas. Connecting a UHD television set to this kind of home broadband network can place very high demands on a base station that may be serving a large radius.

The problem is that mobile broadband is offered on a "best efforts" basis to consumers. All users at a particular base station are competing for the limited resources of the wireless link and stand an equal chance of getting enough to do whatever it is they are trying to do. But, if the resource becomes oversubscribed, the useful data throughput drops through the floor, leading to buffering and service interruption.

����tv Research & Development have been working with colleagues to try to characterise these problems and to help them better understand how mobile networks are different from fixed line broadband. A connection drop-out sometimes means that the network is congested, but sometimes the user has just driven through a tunnel and has temporarily lost the mobile signal. If an application can differentiate between failures like these, it can take different actions to recover from them, potentially leading to a better Quality of Experience for the user.

How it works

You're stuck in traffic again, and the car in front is crawling along. Wouldn't it be nice if you could just pull out onto the hard shoulder and drive off into the sunset, leaving the congestion behind? Maybe you could even pay to use the emergency lane whenever you wanted to. Obviously, the price would be set high enough that not everyone could afford to do that, so the hard shoulder wouldn't ever be clogged up like the conventional running lanes everyone else is stuck in.

It’s nice to dream but, just as the hard shoulder is normally reserved for use by the emergency services, many communications regulators around the world, including those in the UK, the European Union and the USA, have Net Neutrality rules banning content providers from paying to prioritise their traffic over that of others. But communications providers are allowed to treat different classes of traffic – such as all voice calls – with a higher priority.

Since 2016, 3GPP members – including the ����tv – have been designing solutions that improve the Quality of Experience for media streaming that takes advantage of the fact that mobile networks provide mechanisms to detect different flows of traffic and apply different Quality of Service (QoS) policies and charging treatments to them. They devised a new architecture called 5G Media Streaming (5GMS) that enables Content Providers like the ����tv, mobile networks and applications to collaborate together to deliver better results to the user. The solution is based on the premise that the application knows what the user is trying to achieve at any given moment. And if it can share this knowledge with the mobile network, the network is better placed to help the application achieve its goals.

The five main features of 5G Media Streaming are these:

• Content Hosting. A Content Provider can tell a mobile network about a Content Delivery Network that it uses to make media content available, or it can request that the mobile network provides this facility on its behalf. This CDN can be deployed either inside or outside the 5G core.

• Dynamic Policies. A Content Provider can configure a set of Policy Templates in the mobile network corresponding to Service Operation Points: different network Quality of Service (QoS) levels at which its service can operate. For example, there could be a Policy Template for High Definition video operating point that specifies a minimum bit rate of 8 Mbit/s. At the start of a streaming session, the content provider's application can then request this operating point from the network and, if it has sufficient capacity, the network can do its best to achieve that network QoS. In the near future, an application will also be able to ask the network to lower the QoS to do a non-real-time data transfer in the background. If many UEs tell the networks about their intentions, the available capacity can be shared out more efficiently.

• Network Assistance. At the start of a media streaming session, an application can ask the network what bit rate it can offer (bit rate recommendation). This allows the application to pick the best bit rate to stream at from the outset, avoiding the need to probe the network at higher and higher bit rates until it fails. In addition, if an application realises during the streaming session that its buffer levels are reaching a dangerous level (too low in the case of media consumption; too high for media contribution), it can request a temporary delivery boost from the network to quickly replenish (consumption case) or drain (contribution) the buffer. This helps to ensure continuity of service.

• Quality of Experience metrics reporting. During a media streaming session, the UE reports the streaming quality that it is experiencing. Today, the analysis might happen lazily after the session is over and a mobile network operator could use the information to improve the network, for example by engineering out bottlenecks. Artificial intelligence and machine learning techniques can be brought to bear as part of the analytics toolbox. In the future, the analysis could happen in real time, with the network dynamically adjusting itself to route traffic around obstructions. Provision has been also been made to expose the metrics reporting information to the content provider, so the analysis needn’t just be done by the mobile network.

• Consumption reporting. During a media streaming session, the UE reports information about what media is being consumed. Subject to configuration by the Content Provider, this can include information about the currently selected bit rate, the location of the UE and even switching between different mobile networks. Again, consumption reporting information is useful in understanding the performance of the network and can be analysed in real time or not. And the consumption reports can be made available to Content Providers for analysis too.

When used together, these features combine to provide better Quality of Experience for media streaming.

3GPP was keen to avoid the need for mobile handset manufacturers to pre-install special software on UEs (referred to as middleware), although this wasn’t ruled out. Ideally, the mechanisms for controlling 5G Media Streaming on UEs should just use standard technologies that can easily be built into an application or background service which the user can install. The protocols for media delivery should follow established standards, such as and , as much as possible, and the media should be encoded using common codecs widely supported in UEs already, such as and .

The delivery of media over the mobile network (the lower part of the figure below) follows a conventional pattern: an Application Server for 5G Media Streaming (5GMS AS) provides functionality broadly similar to a Content Delivery Network (CDN) and this can either be embedded in the mobile network or deployed outside. The 5GMS AS then offers media streaming services to a Media Stream Handler running on the UE. This is typically a Media Player in the case where media streams are being consumed by the UE; for contribution of media by the UE to the network, it’s a Media Streamer. The Media Stream Handler can be distributed with the application installation package, or it can be installed separately as a background service acting on behalf of several different applications. Eventually, it could even be included in the smartphone middleware.

The 5G Media Streaming architecture takes advantage of a special interface that the Policy and Charging Function (PCF) in the 5G Core exposes to so-called Application Functions. On the diagram above, it’s labelled as reference point N5 and it provides a simple interface that allows any authorised Application Function to:

• Interrogate the current Quality of Service for particular application data flows on the mobile network and receive updates when this QoS changes. This supports the Network Assistance bit rate recommendation feature described earlier.
• Request a Quality of Service treatment for particular application data flows on the mobile network. This supports the Network Assistance delivery boost feature as well as the Dynamic Policies feature.

These features are provided by an Application Function for 5G Media Streaming (5GMS AF) that can be operated either by the mobile network or by a third party, such as a Content Provider. As well as being configured by the Content Provider (marked as reference point M1 in the figure), unusually in this architecture the 5GMS AF also exposes a network interface to a piece of software running on the UE called the Media Session Handler (reference point M5). This is installed by the user, either as part of an application, or else as an additional background service. It offers an Application Programming Interface (API) to applications (at reference point M6) that allows them to launch session handling for a 5G media streaming session using the network interface to the 5GMS AF. The Media Session Handler then interacts with the Media Stream Handler to obtain QoE metrics and consumption information, and it interacts with the 5GMS AF to send reports and take advantage of the QoS features exposed by the PCF that were described above.

Outcomes

It’s one thing writing a technical specification; proving that a technology actually works in practice is another matter entirely. As part of the , ����tv R&D has been contributing to an Open Source reference implementation of the 5G Media Streaming specifications. We have contributed reference implementations of the and the to the effort, and we have also released some as part of our contribution to the project funded by the UK Government to support advanced Use Cases for future mobile networks.

One side benefit of implementing the 5GMS technical specifications is that we have been able to provide constructive feedback to 3GPP on the underlying design. This has led to some evolution of the specifications and the correction of various errors along the way.

What’s next?

As mentioned above, a Background Data Transfer feature is being added to the Dynamic Policies feature in the next revision of the 5G Media Streaming specifications. In addition, support for media contribution from UEs to the network is being improved so that the benefits are equally available in both directions. Finally, the next major revision (3GPP Release 18) will extend support to Real-Time Communication protocols like , in addition to conventional distribution formats like MPEG-DASH.