Many operators are convinced edge computing will improve their 5G business case – almost 70% agreed with that statement ina recent survey of over 80 MNOs conducted by Rethink. But the number with a commercial deployment or clear, near-term plan remains small, indicating the gulf between aspiration and readily available solutions and business models.
Some operators are blazing the trail, however, and will provide valuable examples for others to follow. At several edge-related events over recent months – including the Edge Computing Congress in London last month and numerous edge sessions in this week’s Mobile World Congress Americas – some of these telcos’ efforts have been highlighted. Many of these are US-based, though it is clear that a significant amount of telco edge work is also being driven in Asia, especially China – so far, there is somewhat less commercial detail about those trials, but we expect those details to emerge at next year’s Mobile World Congress, in particular.
One of the notable trends is that most advanced edge deployers are far less religious about the locations of those edge nodes than they were even a year ago. AT&T, Sprint and others are developing platforms that allow third parties to run their apps on telco edge cloud infrastructure – but also that allow the same apps to run across a mixture of telco edges (to achieve wide coverage for services that cover wide geographies or whole metros), and private on-premise edges (for security, ultra-low latency and control).
Here are some examples of operators which are moving the goalposts in real world telco edge models:
AT&T is one of a group of operators which are not only planning edge deployments of their own, but are establishing initiatives and spin-offs to push telco-focused edge cloud to the wider market. Deutsche Telekom has its MobiledgeX spin-off (see separate item), while AT&T has worked a lot through open source efforts, particulary with its Akraino open edge stack.
AT&T says it is concentrating on the middle of the edge to build network intelligence for 5G applications. Josh Goodell, AT&T’s VP of edge solutions, told a recent conference that is is building virtualization and intelligence into the customer edge and network edge, allowing enterprise customers to process data across a mixture of AT&T sites and their own on-premise edge nodes. The telco has trials in factories and hospitals, among others.
But the telco edge is vital too, despite enterprises wanting some applications on their own edge for reasons of control, security and ultra-low latency. “The big difference between this approach and on-premises is that this will span an entire metro, whereas the other approach is centralized to a location,” Goodell said. Network edge compute can support 20ms latency over a 200-300 mile radius, he added.
For its own internal efforts, AT&T has several strategic partners, including Microsoft and HPE. It said earlier this year that it improved network performance by over 50% by deploying Microsoft’s cloud services – which are underpinning much of AT&T’s own migration to the public cloud for its business applications – at the edge of the network. Ofer HaCohen, director of the AT&T Foundry in Israel, said this test alone could make the business case for the telco’s 5G network.
“It opens up so many opportunities and use cases,” HaCohen said. “From a Foundry perspective, one of our key missions is to create synergy and drive innovation by working very closely with the ecosystem around 5G and edge, and to see how these new, cutting-edge technologies can be leveraged to create business value and benefit to our customers.”
The partners started work earlier this year to test the use of Azure to bring edge computing capabilities into AT&T’s 5G network in order to reduce latency and improve performance by bringing cloud services closer to end users. AT&T set up an edge computing environment in its Cloud Foundry in Plano, Texas and also worked with the Israel Foundry to test the network.
The use case which featured in this trial was drone detection and tracking, working with local drone monitoring start-up Vorpal. Vorpal’s software ran on Azure edge cloud, delivered through the Plano test environment, to track four drones.
HaCohen explained: “The discussion was: what if we can take more compute at the edge of the network with low latency and bandwidth? How can we enhance their business? We saw this vast uplift in performance that they can benefit from, specifically in an area — drone monitoring — where low latency is needed for near real-time monitoring.”
The edge environment saw between 40% and 50% better network performance compared to testing in the public central cloud, and speed tests showed latency levels between 24 and 40 milliseconds, with very low variance, which is just as important for services that require consistent quality of service.
HaCohen said AT&T and Microsoft will continue working with Vorpal and other start-ups to test new use cases for the edge cloud/5G combination. “This is just the tip of the spear,” he said. “There’s so much more that can be done, and the more use cases that we create the more customers can see how they can take advantage of these new technologies.”
DT is running edge cloud computing in four of its central offices in Germany with a fifth location planned this year, to support customer trials. In the longer term it will convert
900 locations in Germany into edge cloud centers, or cloudlets, which will host distributed core as well as other virtual network functions (VNFs) for DT’s own network, and could also support customer or partner applications. The operator is using its own MobiledgeX as the platform for its edge applications development and deployment.
DT is working on several applications for consumers and enterprises, including AR gaming with Niantic and a trial with lighting manufacturer Osram for factory and warehouse automation.
SK Telecom has a commercial edge platform in place and will implement 12 edge data centers by 2020. It expects smart factory, VR/AR and video game streaming to be the main applications. It will increase the number of its own edge nodes but also use third party locations.
In March, it unveiled an open edge system which is designed to support a range of commercial services, including a business-to-business platform for enterprises.
SK Telecom says its particular implementation of a mobile edge computing architecture can reduce latency by 60% in 5G applications that require rapid response times, such as robotics control. It calls its platform MEC but it does not fully align with the ETSI MEC (Multi-access Edge Computing) architecture.
The MNO will install micro-data centers at base station and service router locations, and in central offices. It will open up the platform to businesses ‘as a service’, to enable enterprises to launch their own mobile edge services,tailored to their individual models.
To encourage enterprises to make this move, it is releasing an applications programming interface (API) to help them develop their own 5G edge-based services, in areas that require high quality of service, latency reduction and security.
Vodafone is converting some central offices to data centers in most of its territories, though it is taking a cautious approach to offering edge services beyond its own requirements to improve gaming and other consumer offerings. However, its Indian subsidiary, Vodafone Idea, the country’s largest MNO, has a more aggressive approach, given the significant challenge mounted by new entrant Reliance Jio, which has ambitious strategies to become digital service provider and an enterprise operator.
Vodafone Idea (VIL) has established a network-as-a-platform offering to support industries across India, and is working with virtualized mobile network specialist Mavenir to enhance this platform. Initially, the companies have focused on Voice over LTE and will now move on to IoT and content delivery network services.
“Vodafone Idea has the largest Edge Cloud deployment in India. Our partnership with Mavenir will enable continued innovation in optimizing data path algorithms for VoLTE. We look forward to setting up a joint lab and an agile team that will enable on boarding third party applications on our distributed cloud”, said Vishant Vora, CTO of VIL.
Three years ago, Sprint started to deploy a distributed virtualized 5G-ready core network specifically for IoT services. This is now implemented, along with the cloud infrastructure to support it, in 150 points of presence, and the platform will also support third party edge applications. In addition, Sprint will enable third parties to run applications over a combination of their own and Sprint locations.
“You can’t build a network and bring services to it. You have to see where demand is and take the network to that, Parvin Mahajan, head of customer operations for Sprint’s IoT business, said.
US provider CenturyLink recently announced plans to deploy edge computing to support its own NFV (network functions virtualization) program, and it will leverage the same infrastructure for edge cloud services.
CenturyLink has pledged to make investments worth “several hundred million dollars” in edge compute infrastructure and services following a successful US-based trial. It will kick off by building 100 new edge sites, with a target of focusing on applications which require latency of 5ms or lower. It plans to offer a range of offerings including infrastructure-as-a-service and managed services.
The plan will not be confined to the USA, if the business model proves workable. CenturyLink has a global network of 450,000 route miles of fiber, connecting over 2,200 public and private data centers and over 150,000 on-net enterprise buildings; and it is extending network colocation deals in many markets to increase access to infrastructure and allow customers to run distributed cloud workloads close to the edge of the network.
Dave Shacochis, VP of hybrid IT product management, told analysts at HeavyReading that he sees the strongest demand in the retail, hospitality, food services, logistics, transportation and manufacturing sectors – all industries which operate over very wide geographies. Here, the telco says it can save companies from building out large numbers of their own edge data centers – it gives the example of a customer which had been planning to deploy edge servers in each of its 2,000 sites in the USA to support a new business application, but will instead deploy the application on CenturyLink’s 100 edge nodes without losing the 5ms latency required.