Edge computing is at the peak of its hype cycle, at least in the telco market. Mobile network operators (MNOs) quote edge as the technology that will make their 5G business case fly, as frequently as, last year, they were citing the Internet of Things, or the year before, self-driving cars. In many cases the edge – deploying compute, storage and connectivity close to the users – is closely tied to the IoT in operators’ strategies.
But there are clear challenges in this approach – neither edge nor IoT is mature in terms of platforms, business case or clear demand; and while both will be highly significant in changing the mobile network model in future, neither is MNO-specific. At the edge, as in the IoT, operators will have to find their role in a complex ecosystem, and try to ensure they are more than 5G or converged bit-pipes for all the data passing from sensors to edge nodes to the cloud.
Some operators are putting together clear and rational roadmaps, though in most cases actual deployment is at the earliest stage and the return on investment (ROI) model is not fully clear.
They are battling with the gulf between early assumptions about the telco edge and the emerging reality. The early operator interest in edge, as encapsulated in the first ETSI MEC (Multi-access Edge Computing) architectures. This assumed that existing operator locations – central offices and cell sites – would be enough to support an edge for nearly all purposes. This would include the MNO’s own services, such as enhanced AR-enabled gaming (probably the only well-defined and scoped telco-driven edge application so far); as well as support for many third party enterprise and industrial applications, especially those in the IoT which required localized capacity and connectivity to reduce latency, support huge numbers of devices, and protect privacy.
In reality, while the demand for edge computing is growing far more quickly in the enterprise than classic telco, consumer-facing services, in many cases these industrial players want on-premise edge cloud nodes; or they will use public edge networks, but the sites will be provided not just by operators but other deployers like webscalers.
So there are question marks over how far the telcos really can leverage their existing sites to provide an edge network with the right number and location of nodes for many enterprise and IoT purposes, not just to enhance their own services for their subscribers. This is seen in the hesitancy that some tower operators are now exhibiting, even though they were initially, especially in the USA, seen as natural organizations to deploy edge clouds on their sites. On one hand, there are new start-ups, such as EdgePresence, following the pioneer, Vapor.io, into the tower-oriented neutral host edge business. On the other, Crown Castle, the most optimistic about this model, seems to be pulling back a bit from its commitment to edge.
Some operators, such as Sprint, SK Telecom and Vodafone Idea, are setting out detailed edge strategies with different commercial objectives, while others, like BT, remain convinced that just a few edge nodes per market will be required, and that the monetization opportunities for operators will be limited; and others point out the technical challenges of distributing the cloud in tandem with distributing the RAN and core, which for many promises to be the optimal approach in order to leverage 5G roll-outs as well as possible.
It is clear that the greatest power will lie with companies that can aggregate many edge locations and provide dynamic orchestration of services – and in future, microservices – across those edges, moving towards network slicing. Companies which are seen as enablers of this are hot property in terms of venture capital and acquisitions, and a mini-bubble at the edge can be expected. In the past week or so, Intel has acquired Pivot while EdgeConnex is said to be in play.
Perhaps the biggest driver to deploy this complex edge network will come when operators start to roll out a more distributed core network. The pendulum is swinging away from highly centralized 5G core architectures and many MNOs are planning more distributed build-outs when they come to implement a cloud-native, sliceable core. This will necessitate turning central offices and even more distributed locations into micro-data centers to support the core network functions, and sometimes virtualized RAN basebands too. In that environment, it makes sense to use the cloud capacity to support applications too, and organizations like Ericsson’s Edge Gravity are pushing use cases and developer platforms that will take advantage of this.
Deutsche Telekom, for instance, is implementing its 5G cloud-native core in parallel with converting 900 of its telco 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. So far, central offices in Berlin, Frankfurt, Hamburg and Munich have been turned into cloudlets, and Dusseldorf will be next. The operator is using MobiledgeX, a start-up backed by DT, as the platform for its edge applications development and deployment.
But to achieve the coveted orchestrator role in the edge will be a target for some edge-focused operators, but also for enterprise specialists, webscale providers and others. It will be vital for the MNOs to start planning now for how to make maximum use of their assets – sites and 5G – while signing up partners and devising processes that will enable them to take a bigger role in the value chain.