Tier 1 telcos have struggled to ignite much enthusiasm for the mobile edge network, despite the obvious potential for applications that require ultra-low latency or that generate huge amounts of data most economically stored and processed locally. One problem is that most of the proposed applications have not been deployed yet, making the mobile edge look like a solution chasing a problem that may not exist, or that at any rate has not yet become pressing.
Vodafone is one of the most active operators with a mobile edge program. It has been unfolding over the last few years through various partnerships and initiatives, with AWS Wavelength playing a crucial role as the source of edge compute, and storage capabilities inside its networks.
To some extent this problem igniting the Multi-Access Edge Computing (MEC) field is acknowledged by Vodafone itself with its statement that the key to MEC services is “the ability to bring to life ideas that would otherwise have been impossible.”
The challenge is converting the “now possible” into the “must have,” which is why Vodafone is now promoting five use cases as primary targets for evangelistic marketing campaigns – as well as trials and demonstrations.
These are mixed reality, video analytics and streaming, live video production, V2X (Vehicle to Everything) communication, and AI in various guises, almost invariably some flavor of machine learning (ML).
This list is curious, in that it ranges from vertical sectors to enabling technologies. For the former, the automotive industry is touched on by V2X, relating specifically to communication between vehicles or with roadside infrastructure. The other vertical example belongs to the broadcast industry, with the video production and delivery angles.
For the technologies, mixed reality via AR and VR is more self-evident, but AI is the most nebulous and ill-defined. The latter potentially includes both existing deliverable applications, processes that enhance network performance or efficiency, and new analytics applications that can only be executed at the edge because of the low latency required.
Existing applications that involve the edge only because that is the point of access are not of particular interest here.
However, some lessons have already been learned by the MNOs that are hoping to monetize their network infrastructure to expand into edge and MEC services.
Among major telcos Verizon is the one to exhibit the most public agonies over MEC. When presenting its financial outlook for 2023 in January the company downgraded expectations with CFO Matt Ellis singling out slower than expected adoption of mobile edge compute and private 5G as the major factor for the revision.
This experience has been shared to varying degrees by other telcos around the world, but some such as Vodafone remain optimistic that impending acceleration in the field has merely been postponed, and that now is the time to push harder than ever.
In contrast, Deutsche Telekom decided to focus on edge computing management through a specialist subsidiary called MobiledgeX launched in early 2018 and became involved in some standardization activity with the GSMA. Yet in the end DT backed off and sold the company in April 2022 to Google, whose main motive seemed to be to add the acquired code to its collection of software it had made available open source, and effectively shut the brand down.
As Vodafone has come under investor scrutiny, it has undergone leadership shakeups. In addition, it is still in merger discussions with rival Three UK, which Three is describing as an existential concern. These talks are thought to be nearly finalized. To this end, it cannot afford to launch blindly into a supposedly promising new vertical, before discovering that there is no money in it, and so examining its strategy is prudent.
Under the MEC umbrella, two AI-based tools are being put to use to enhance network operations. These are deep reinforcement learning techniques and federated learning frameworks, and are being put to use by routing data traffic more efficiently, or turning down power supplied to some components at off peak times to reduce energy consumption.
This category of edgy AI scores directly by reducing costs rather than generating new revenues, although there is scope for exploiting some of the applications by offering them to other telcos, as well as enterprises as part of private 5G services. That is one reason some telcos have launched dedicated MEC units, even if some have floundered.
The likes of Vodafone should be able to realize operational improvements in their networks using such MEC tools, even if they are not particularly successful in gaining external customers.
Another category of edge AI is involved in latency-sensitive applications that are not possible using a centralized networking approach. This overlaps with some of the other MEC monetization categories targeted by Vodafone, especially mixed reality in the industrial sphere. We do not anticipate a significant role for mixed reality deployed at the edge for consumer applications in the near future, but it is already starting to figure in manufacturing and engineering for applications of digital twinning.
An example of this is found in industrial and automotive giant Bosch, which already uses digital twins as software-based replicas of components such as electrical motors to analyze sensor data in real time for predictive maintenance in the short term and ongoing optimization longer term.
MEC allows processing to be performed locally to enable the impact of changing sensor data to be calculated quickly in software within the digital twin model. The old alternative involved sending data off to a centralized process, and relying on engineers to inspect these repositories in a fairly manual fashion.
Of Vodafone’s other potential MEC revenue candidates, the video-based examples seem quite promising. On the production side, MEC enables cost and efficiency savings by optimizing the workflow – spanning from video capture, to editing, and the initial contribution feed to the broadcast networks.
The key point of MEC is that it enables the optimum division of computational labor, taking account of factors such as latency, bandwidth and location of resources. In many cases the edge will be the end device itself, whether a smartphone, drone, or other IoT component.
Vodafone uses the ETSI-defined term Multi-access Edge Computing to define such approaches where computation and storage capabilities, at one time confined to centralized data centers, are distributed increasingly across a multi-layered hierarchical approach to smaller local centers and end devices themselves.
The company has claimed several MEC milestones, such as being the first in Europe to launch distributed MEC services in 2021, aligned with its cellular infrastructure through AWS Wavelength to serve MEC capability in principle anywhere reached by its network. Further developments included the Edge Innovation Lab in MediaCity, Salford, UK, opened in November 2022, this time with the claim of a country first of its kind, offered to customers, partners, academics and software developers for evaluating potential MEC applications.
Vodafone has also pitched its V2X platform to telcos, but here has been struggling for traction so far. V2X has suffered to some extent from the failure of autonomous driving to gain traction and take off beyond niche deployments away from public roads in campuses or ports for example, largely because of the friction imposed by legacy and the need to demonstrate almost flawless safety.
The irony here is that while autonomous driving has been sold in large part on the promise of much greater safety, given that over 95% of traffic accidents are the result of human driver error, it has been derailed by several high-profile human-error accidents.
Car connectivity itself over cellular networks is booming and serving infotainment, navigation and remote diagnostics, but the more local V2X element that was supposed to usher in autonomous driving and more advanced safety through automated emergency braking for example, has only arrived in piecemeal fashion.
A major source of frustration for the autonomous movement is the almost indefinite transition phase towards autonomous driving during which human drivers still have to be alert and capable of taking control at short notice, therefore not gaining most of the benefits. Even more decisively, fully equipped autonomous vehicles would have to coexist with older ones lacking even basic capabilities such as cruise control, which would mitigate the gains in safety and make accidents resulting from system level failures more likely.
In practice, self-driving vehicles will need to be fully autonomous in the true sense – capable of making all decisions such as braking and steering in isolation, without external assistance via V2X. It is true there is the concept of a layered approach where such assistance adds value when it is most needed in congested road conditions, yet the direction of travel may be towards the vehicle itself becoming the edge where essential computation takes place, with less dependence on V2X connectivity.
Vodafone is an interesting case with its origins as a national UK operator that harbored ambitions of becoming a multinational media and communications giant building out into TV across multiple countries, followed by expansion in IoT.
To some extent it achieved the first objective for a time especially in Europe with forays into video and fixed line broadband, but has struggled to escape the telco box, as reflected in its stock price which amid oscillations has remained broadly unchanged for two decades, having peaked in 2000.
MEC is the latest attempt to resume expansion, and certainly Vodafone has been pushing harder than many rivals, some of which have become impatient over lack of growth in the field.