Challenges for Mobile Entertainment on 4G
Consuming entertainment on mobile devices is not new. In fact, 4G and LTE networks already provide broad connectivity and access to streaming content. These are typically delivered via public Content Delivery Networks (CDN) with Edge Caches that are peered with mobile Communication Service Provider (CSP) networks.
However, as mobile content consumption has become more and more popular, it has exposed numerous scalability challenges in the current setup. Streaming is bandwidth intensive and competes directly for the limited capacity shared between applications already using the mobile network. In addition, the “appliance” based nature of 4G infrastructure (i.e. Radio Backhaul and Core networks) prevents the deployment of caching infrastructure deep within the CSP infrastructure, requiring additional investment to support per session data being pulled from the Edge Caches. This is a major concern for CSP’s who, at least to date, have been unable to recoup any expenditure by negotiating even a small percentage of the revenue generated by OTT services.
The consumer experience also leaves a lot to be desired. Most mobile content is distributed in SD quality due to network limitations, not to mention additional challenges such as re-buffering and delayed start times. Video performance also may suffer due to network congestion, high latency and packet loss. While this may not present a large challenge to casual users consuming video soundbites on the go, it presents a huge problem for millennials and others who are eschewing traditional Fixed Access for Mobile Broadband services.
In this blog we will summarize some the key developments in 5G networks. We will take a closer look at how 5G is aiming to enhance the mobile streaming experience and resolve some of the scaling and quality challenges it is currently facing. We will also look at some network considerations that service providers should consider when deploying a 5G-based streaming service.
5G – The Promise
5G New Radio (NR), the fifth-generation technology standard for broadband cellular networks, promises to address some of the challenges inherent in 4G mobile streaming. It implements unique characteristics and architecture that aim to enhance the mobile streaming experience both at home and on the go, and is intended to help service providers to meet the demands of their consumers in a cost-effective way.
5G’s promise includes its high bandwidth transmission which allows HD and even UHD streaming on mobile networks. This alone has the potential to replace fixed access services with mobile broadband. It supports high capacity which allows improved scalability for distribution of popular content and large-scale live events at a reasonable cost, using a minimal number of server sources.
5G also supports a unique caching architecture which allows a reduction of the streaming latency and a reduction in the bandwidth requirements by distributing caches throughout the CSP network. 5G also allows the user access to adjacent computing resources, which can efficiently enable real-time cloud gaming. The 5G architecture may also encourage CSPs to further scale their streaming content distribution as it provides them with a financial incentive to do so.
5G also allows introduction of mdelivery and network slicing which can further reduce streaming latency and allow better management of the overall network resources. We’ll talk more about these later in the post.
In the next several sections we will elaborate further on the unique characteristics and infrastructure of the 5G network.
5G Supports Higher Capacity for Streaming Media
One of the most obvious 5G improvements will be manifest in the mobile network and the consumer-facing aspects of 5G service.
5G Radio Access Networks (RAN) are fast, dense, and lag-free, and streaming services benefit greatly from additional network capacity and low network latencies. 5G offers up to 10 Gbps of shared bandwidth, scales to 10 Tbps/km² of network capacity and can reduce overall end-to-end network latencies to less than 5ms.
The 5G spectrum covers three specific wavelengths, each having very different attributes and support for well-defined streaming use cases. “Low-Band” utilizes the sub-1 GHz spectrum and has the distinct advantage of maximizing network coverage. Its signal propagation is far-reaching and is largely unimpeded by walls, glass and other objects. This means that a consistent experience can be sustained for streaming, even while the consumer is on the move, but only if the network is unsaturated since capacity is limited.
“Mid-Band” utilizes sub-6 GHz spectrum, providing speeds of up to up to 1 Gbps with extreme low latency. This is the sweet spot for streaming applications, particularly those that are immersive or require high levels of interactivity.
“High-Band” or millimeter wave (mmWave) operates at frequencies greater than 24 GHz. It delivers Gigabit speeds and is especially well-suited for content delivery. Its main disadvantage is that it is hindered by extremely low signal propagation combined with poor object and building penetration. To overcome this, mmWave deployments consist of many ultra-small, ultra-close cells arranged in a mesh that cover major urban areas. These resemble WiFi substructures rather than traditional wireless networks and provide connectivity from antennas deployed approximately every 250 meters. It is best suited to providing a replacement for broadband services in urban and suburban neighborhoods and provides stiff competition to existing Fixed Access infrastructure.
To take full advantage of 5G infrastructure deployments, the video caching model is also changing and is migrating deep within the CSP network. This transformation is possible as 5G avoids the predominantly single vendor, appliance-based ASIC’s solutions that powered 4G and LTE networks.
5G networks are virtualized, cloud native platforms that connect the Radio Access Network (RAN) to the operator 5G Core (5GC) and can support public or private commodity cloud compute and storage functions. These functions can be deployed either adjacent to the 5GC or as a function within the 5G network itself. This new compute model is already being implemented by Public Cloud Providers who are deploying Multi-access Edge Compute (MEC) instances within the CSP network. These near-edge deployments are positioned in proximity to the viewer and eradicate the 100 to 200ms round trip times to the peered public CDN’s or regional cloud services currently required to retrieve content fragments.
This provides more available and more robust connectivity, and when combined with new protocols like HTTP/3 and QUIC that offer 0-RTT session connections, end to end streaming workloads can now potentially support single digit millisecond responses.
This allows CSP’s to host services on behalf of Public CDN’s or expose capacity on their own CDN infrastructure that can be leveraged by Content Owners and OTT Service providers. They can begin to monetize their investment in 5G by bringing content services into their networks and offering CDN at the edge, allowing them to participate in the streaming value chain instead of being a passive member.
What’s Your Edge?
Edge Zones provided by Public Cloud Providers deliver the highest tier of compute and storage within the CSP network and this infrastructure is being deployed above or as an adjunct to the 5G Core. While not strictly part of the 5G Edge, these “public” services are ideally positioned to support latency sensitive workloads that include Edge Caching and Content Delivery. Infrastructure as a Service (IaaS) and Platform as a Service (PaaS) is supported with the latter exposing common “big data” cloud services that can be shared between streaming workloads. For CDN vendors that are already cloud-native, workload deployment is a simple “lift and shift” operation.
Deployment on ETSI Multi-edge Access Compute (MEC) is also a consideration but slightly more involved. MEC is a highly localized compute platform deployed as an integral part of 5G. MEC allows workloads to be deployed within the 5G network and exposes real-time, actionable data related to RAN conditions and user equipment locations. This data can be further leveraged by an Edge Cache to optimize or personalize the user experience.
A key challenge is that ETSI MEC is typically deployed in the CSP’s private address space. Security must be a primary consideration when bridging to public networks, something that has already been commoditized by AWS within their Edge Zones.
Operator 5G Network and Cloud
Compute and storage deployed within the 5G network is best suited for applications that provide a high degree of interactivity and personalization. For streaming workloads that are more general in nature such as longtail or best effort content delivery that require more compute power, or can tolerate some latency, regional cloud services or public edge networks that are peered with an operator are generally a more suitable location for deployment.
Other Network Considerations: TCP and Volatile Network Conditions
The unique 5G characteristics combined with the expected mobile streaming developments raise a challenge in the ability to fully utilize the 5G delivery on dynamic or congested network and to allow the best possible quality of experience for streaming services
Transmission Control Protocol (TCP) was originally designed for Fixed Access networks and infrastructure and can often perform poorly when presented with the unique characteristics of mobile networks, even 5G.
Bandwidth available to a viewer can vary over time based on location and utilization as devices connect to and leave the network and adapting Video Quality (VQ) bitrates to current network conditions is challenging. Sending content too slowly results in low quality renditions being delivered and inefficient use of available bandwidth, whereas sending too fast can lead to long packet delays and packet losses.
Deep buffers are implemented in mobile networks to contend with massive bandwidth variability and can also generate problems for streaming services. TCP only decreases its sending rate after encountering packet loss and thus causes high buffer occupancy resulting in packets experiencing excessively high latency. Non-congestion packet losses where a user is handed off from one base station to another are also erroneously interpreted by traditional TCP as network congestion.
TCP's traditional control algorithms which reduce the rate when packet loss and congestion is detected often fail to gracefully adapt to such volatile network conditions. The enhanced deployment of latency-sensitive live video services together with the burstiness of other 5G traffic (e.g. IoT) is expected to aggravate these problems.
To address some of the challenges, an alternative approach was proposed - Bottleneck Bandwidth and RTT (BBR). BBR models the end-to-end pipe as a single link, repeatedly probing the bandwidth and round-trip-time (RTT), pacing its sending rate to the link’s bandwidth over time.
Though BBR has the potential to perform better than the built-in TCP Cubic, research by Tomoaki Kanaya et al wasn’t able to conclude that BBR has the clear performance advantage. There were even certain cases (e.g. on “Mixed Band”, 1Gb network) where TCP Cubic achieved higher throughput than BBR.
One of the challenges of the current congestion control algorithms is their ability to react efficiently in networks with large buffers, extremely low latency, and high jitter networks such as the 5G network. An additional promising alternative is Performance-oriented Congestion Control (PCC).
PCC embraces a different approach and treats the network as a black box. It constantly observes the implications of sending at different rates and leverages a machine learning algorithm to adapt the sending rate in the direction that yields better performance. Recently, PCC was deployed on an operational 5G network, and the preliminary results show that such dynamic congestion control is able to better adapt to the unique dynamic of the 5G mobile networks than the traditional methods.
Advanced Features for Stream Optimization
Massive MIMO
Delivery of streaming services needs to be able to scale to thousands of devices from a single location, and 5G can increase the volume of mobile network connections by more than twenty times. This feature is referred to as Massive Multiple-Input and Multiple-Output (MIMO). While MIMO is not new and was deployed with 4G technology, 5G delivers immense exponential capacity by providing dozens of antennas on a single array.
Beamforming
While massive MIMO expands network capacity, it also has the potential to create cross interference due to the density of cellular traffic. A technology known as beamforming provides a solution for this as it makes more efficient use of the available spectrum by identifying the optimal over-the-air route to a device and reduces interference for nearby users. Packet movements and arrival times are carefully orchestrated, eliminating interference and driving data rates, and the technique is particularly useful for mmWave deployments as it helps avoid objects that could potentially block signals, by concentrating the beam directly at the user device.
Full-Duplex
5G also provides a full-duplex model for greater efficiency. 4G networks take turns transmitting and receiving over the same frequency or use two separate frequencies. This is inefficient. 5G transceivers operate in full-duplex mode and can transmit and receive at the same time on the same frequency.
Network Slicing
While all these features can improve the QoE of premium streaming services, perhaps the most interesting feature is Network Slicing. CSP’s can assign virtual dedicated network slices within a common RAN infrastructure and offer a different class of service for media consumption while allowing consumers using less “experience-critical” applications to continue with best effort download speeds and latencies. For a CSP or Content Owner, this is a revenue-generating opportunity, as slicing can be implemented as a billable service.
Entertainment Related Use Cases for 5G
Live and on-demand streaming services are the most obvious, and potentially one of the “killer” use cases for 5G infrastructure. While 5G can have a huge impact on overall video streaming quality, it also helps democratize content delivery services, particularly for millennials or consumers in developing countries that consume most of their services over the mobile network. With over 10 billion active mobile device connections globally, 5G has the potential to bring content to everyone.
5G also has the ability to extend the entertainment concept to an immersive, interactive, and tactile shared experience for video entertainment, cloud gaming, and Extended Reality (XR). For devices with limited compute resources, 5G makes off-loading processing to the CSP edge a reality and allows the network to be “bookended” with shared functions that reside both on the device and in the operator edge cloud and collaborate between the two. Leveraging 5G’s low latency connectivity in this way means that the majority of XR workloads can be executed in the 5G Edge that is more suited to this type of heavy lifting, while minimal fine-tuning and adjustments are done on the device prior to rendering.
5G will also allow the introduction of emerging streaming-based services including wider usage of virtual healthcare, real-time surveillance technology, and efficient information delivery to, and between, autonomous vehicles.
Finally, services and business models that depend on personalization that include hyper-targeted advertising can be based on highly accurate, pinpoint location data and this will help drive additional monetization opportunities.
Conclusions: Adopting a Holistic Approach to 5G
In 2019, media interest and publicity created a lot of buzz as early adopters began to roll out 5G Networks. Inflated consumer expectations led to disappointment and cynicism as coverage was spotty, even in dense metropolitan areas, and device vendors did not have the user equipment to take advantage of what 5G there actually was.
Today, most major global CSPs are actively deploying and scaling their 5G services. Widespread implementation of 5G technology is accelerating, as are the sales of devices like the iPhone that can take advantage of these new networks, although it will still be a long time before 5G provides 100% nation-wide services in most countries.
According to the GSA, (Global Mobile Suppliers Association) 5G is being deployed globally these days. As of December 2021, 200 operators in 78 countries and territories had already announced their 5G service launches (mobile or Fixed Wireless Access).
As outlined above, 5G offers high bandwidth networks, low latency, and enhanced performance for streaming services, yet it still faces challenges with the continued use of inherited, legacy technologies, the TCP protocol stack and congestion control algorithms. Fortunately, 5G has evolved lock step with new network protocols including QUIC, Performance-oriented Congestion Control (PCC), and others. Streaming media trials have already shown that adopting a holistic approach instead of focusing on a single technology silo leads to exponential improvements in the overall customer experience.
The rollout of 5G services has required a massive investment in capital and personnel. CSP’s are facing the prospect of a trillion dollar-plus investment and need a greater ROI than provided by consumer cellular services alone. A case can therefore be made for a more aggressive rollout of enterprise applications, particularly those related to streaming media and entertainment to match 5G consumer cellular growth, and cover CSP investments.
About Compira Labs
Compira Labs is a pioneer in ML-powered content delivery. Its software-only solution, which includes PCC congestion control, can easily upgrade any CDN to deliver best-in-class QoE even in the most challenging last-mile networks. Compira Labs’ solution can dramatically improve user experience for latency-sensitive and bandwidth-hungry Internet services such as live and on-demand video streaming, game downloads etc. It is available for both HTTP/TCP and HTTP/3/QUIC delivery methods.
Glossary
Beamforming - a traffic-signaling system for cellular base stations that identifies the most efficient data-delivery route to a particular user, and it reduces interference for nearby users in the process.
Backhaul - network infrastructure responsible for transporting communication data from end users or nodes to the central network or infrastructure and vice versa.
Edge Zones - small-footprint extensions of the public cloud that are deployed on a customer premise. Such distributed public cloud offerings enable low latency access to localized computing and storage.
Multi-access edge computing (MEC) - a network architecture concept that enables cloud computing capabilities and an IT service environment at the edge of the cellular network
Peer cache - a built-in configuration manager solution that enables clients or service providers to share content with other clients directly from their local cache
Radio network – a network providing a seamless connection with a local area network (LAN) via radio waves instead of wires. E.g. Wi-Fi, 4G network
Radio Access Network (RAN) - the part of a mobile network that connects end-user devices, like smartphones, to the cloud.