Mobile World Congress was awash with edge demos that demonstrated the potential to use compute at the network edge to deliver low latency applications. It feels like we have been here for some years now – ever since Nokia and Vodafone started talking about Liquid Apps and using edge analytics for road traffic management in 2013.
What the market is still waiting for is the deployment and commercial models to come together to place applications in a mobile network edge – moving from vision and proof of concepts to delivery.
Although the mobile access network is not the only “edge” that could be used to deliver edge applications, mobile operators believe their bit of the network edge comes with certain advantages. First, and most obviously, it provides secure, authenticated connectivity with appropriate QoS and policies applied to mobile devices (and other devices that are peripheral to the SIM device) over the air.
Secondly, mobile operators believe they have the real estate and the backhaul connectivity to provide the required amount of edge node coverage – ie the amount of territory that lies within the latency budget of an edge node – that developers will require. An added element of this is that unlike on-premise edge or private edge networks, the mobile edge offers a portability of function that fits certain use cases where the use group is mobile.
Third, as they move to cloud and virtual RAN and to distributing elements of the core network to get certain functions closer to end users, operators may already have a potential ROI case to install the racks, power and so on to underpin that edge compute. The RFPs for vRAN are filtering through and there’s increasing activity in the hardware sector, which is delivering an increasing number of edge-optimised server variations to the market. That means that operators are getting the means to deploy “edge compute” infrastructure to act as public cloud providers offering content owners and app developers a distributed cloud presence that can be dynamic, portable and also come with that much closer presence to the user.
That’s the theory, but the drive to open up the edge as a distributed cloud hosting platform has presented operators with two key challenges – one that they know they have failed before, and one that they haven’t yet faced.
The first challenge is to partner with and sell to application developers, web companies and content providers. It’s a huge generalisation, of course, but broadly speaking operators have a terrible name with the web and applications developer community. Sure, things have changed, but it’s still going to be difficult for operators to set up cloud practices that can engage with developers in a way that matches client expectations of public cloud providers. There is also the small matter of building the management and control plane of distributed cloud workloads across the network.
These two challenges, one cultural and process-drive and one technical, may create the role for an intermediary that manages the operator edge cloud environment – acting almost like a network asset aggregator.
Operators have attempted to do stuff like that before and they haven’t succeeded
Enter the edge aggregator
One very current example is MobiledgeX, the independent edge computing provider created by Deutsche Telekom that provided the platform for DT and for SKT’s edge demos at MWC 2019.
MobiledgeX has just announced the first public deployment of public edge cloud deployments that use its R1.0 edge cloud. CTO Sunay Tripathi says that the company provides all the cloud management and control, thereby bringing customers to mobile operators. Because it acts as a trusted provider with the MNO, it has hooks into users’ location and authentication. That means that MobiledgeX takes care of deploying the app close to the users where and when they require the app.
“We protect the developer workflow in the public cloud, [deliver] all containerised workloads and once the containers are ready they onboard with us. And what we do is based on where the users are – say you are running a game with maybe ten players, for the duration of the game we will bring that application close to the gamers”
“Because we are a trusted provider we can integrate with the operator back end, so when the user connects we automatically know what service to use, what game to use, and we also know the nearest edge location based on where you we are, and we dynamically deploy a container backend there to provide service.”
“So two things are achieved: one we optimally use the edge location but the developer doesn’t face the burden of where knowing where things need to be. And as it is on demand the users get a much richer experience.”
Tripathi says that applications that might require this sort of dynamic edge presence include “anything that requires an immersive experience; it’s not just gaming, it could be multiple robots in a factory, or a bunch of workers wearing Hololens for AR use cases, or a bunch of drones swarming together. All of these thing create a short term location based group with an immersive application that needs heavy compute.”
The model means that MobiledgeX brings customers to the operator. MobiledgeX aggregates operator resources and pays operators for the resources it uses. It then charges developers for their utilisation of those.
“Operators are friendly to us because we are bringing them revenue and there’s no barrier to entry – they can start making money off use cases today.”
In the US where there is a different model of tower ownership MobiledgeX is working within CrownCastle, and deployments are starting with KineticEdge. “Based on what we have achieved we have started to talk to Asian operators as well,” Tripathi said.
Right now the idea of shifting workloads around a very distributed range of edge locations remains something for the future
Does Tripathi think that operators will attempt to go direct to the market with edge plays?
“The interesting part is that operators have attempted to do stuff like that before and they haven’t succeeded. It requires some innovation, invention, cutting edge people, that’s why we are based in Silicon Valley.”
Also, Tripathi said, developers don’t want to deal with deploying in a different manner for different operators – either within the same country or across borders. The aggregator model means that they only need to design and deploy in one way, and users with different MNO providers can access the same applications.
Tripathi added that he thinks that DT will use MobiledgeX as the control plane for its own edge deployments. “Because bear in mind the software to do all this is not easy, a lot doesn’t exist. And we also made sure we’ll open source all the critical pieces, so that gets all the collaboration going because the operators are not threatened by us, they want to participate with us and work with us.”
Right now the idea of shifting workloads around a very distributed range of edge locations remains something for the future. MobiledgeX has presence in the regional offices of DT in Berlin, Munich, Frankfurt and Hamburg, and can service each entire city from there. “Typically at each location we can deal with anything from a few servers to multiple racks – depending on what the operators have. We are trying to use the existing infrastructure they already have as part of 5G and NFV transformation.”
Tripathi said that for LTE edge applications these points of presence are sufficient to meet latency budgets – but that will change.
“Right now in Europe the latency difference between a Central Office [this is the more distributed location in telco jargon] and a Regional Office is not that much. It’s really a few milliseconds. But there’s a huge difference between public cloud and a central or regional office.
“But as these games, robots etc increase and we start getting more 4k video and collaborative video, then we will need every millisecond. When 5G starts coming out application developers will want to get every millisecond, so we will deal with more real-time critical workloads. In those cases we will want to get closer to towers and Central Offices, but currently a regional deployment is good enough to service a City.”