Close to the edge

From AI-based facial recognition, to driver monitoring to high level cloud gaming - the edge demos at MWC2018.

One discussion that took place this year within the “other” MWC – the one that takes place amongst those who attend to find out about new phones and devices, rather than network technology – was about how little innovation we have seen in device form factor and function.

Well, there’s one possible source of innovation that they missed – the network itself. The dumb client isn’t new, but what if mobile devices accessed applications running as cloud instances in the network. These would be sited at the edge, providing for low latency and avoiding a heavy transmission link to the centralised core. You could free up device power and space, or use existing onboard processors and power for other uses. Could that bring about disruption in the device space – new form factors and capabilities?

InterDigital’s Alan Carlton made this argument, perhaps more as a theory than a deeply held conviction, at a panel session that rather artificially pitted core network versus edge investments.

But even if Carlton’s proposition is as yet far-off, there’s no doubt that edge deployments are closer. This is a result of several investment drivers, some interconnected.

First – the need to relieve cell and backhaul congestion by dealing with traffic locally, and/or caching content. Second, the thinking about vRAN splits meaning that a certain number of points of presence will be required in the network to host vBBU instances – providing more affordable real estate for edge compute platforms. Third, the potential for edge platforms to host apps that can benefit from cell level intelligence and/or local breakout – providing new revenue streams. Fourth, IoT capacity and low latency requirements driving the disaggregation of control and user planes, and the distribution of core network functions through the network – again this presents an opportunity to integrate edge compute. Fifth – 5G low latency and high capacity use cases that could benefit from edge infrastructure, such as connected (not necessarily autonomous) cars, VR and AR. Sixth – competitive pressures to roll out as enterprises ponder private networks and the use of unlicensed spectrum technologies.

There was enough buzz about edge before the event for TMN to make it the focus of one of its preview articles. Indeed HP made the Intelligent Edge the over-arching tagline on its booth, which for a company that has been involved in cloudification projects and NFV since the start, is a sign either of the current direction of the telco cloud, or of HP’s own current market status. Either way – worth noting.

The ability to use an edge cloud location to host high speed, processor intensive activity, was being displayed on a number of booths.


Vasona was showing high speed multi user gaming, working with LiquidSky as the cloud gaming partner. The point of the demo was to show the sort of responsiveness and excellent graphics rendering you would expect of a console-based game – but being delivered to your phone over a mobile network connection.

The processing and rendering was being carried out not on the device, but in a gaming server located at Vasona’s SmartAir edge platform.

CTO Rui Frazao  said that the company is working on a potential commercial deployment with one operator that has a natural fit between its customer base and gaming. Interestingly much of the push has come from the operator marketing department.

Frazao said that operators are still making initial edge investment cases around capacity relief and traffic optimisation, but that they are also looking to future proof those investments by looking for new business cases. For Frazao, the “sweet spot” for the edge is the Central Office, which although it sounds very central is in fact the “one step back” location in the network that can give visibility to hundreds of site sectors at a time.  


On its booth Samsung Networks was showing something very similar to Vasona’s cloud gaming demo – a “console in the mobile edge cloud” demo with two gamers playing on phones connected to that edge cloud over a (Samsung) wireless link.


Saguna had teamed up with Vodafone and Amazon Web Services to design a demo that showed a “dumb camera” monitoring a driver.

An AI app at an edge server processed video analytics to determine if the driver was driving in safe mode. If not, it delivered an alert. So in the demo the analytics could determine that the drive was holding a phone, instead of the wheel, for example. Tally Netzer, CMO at Saguna, said that on-board analytics does exist for driver monitoring but it is expensive and can become outdated. An edge based platform would reduce the cost of the camera, and hence the impact of potential damage or loss, as well as enable any new algorithms.

What’s clever here is not just the app itself (although the driver recognition smarts is cool) but that the demo incorporated a public cloud web services environment – Amazon Greengrass – on Saguna’s MEC gateway. The camera is “dumb”, the smarts are at the edge, being able to use IP source and destination to “break out” the camera traffic to Greengrass for the analytics to take place.  


Intel had a facial recognition demo using edge-based AI on its FlexRAN/MEC solution and live trial deployed in China by ZTE* and China Unicom*

The demo setup was using ZTE 5G radios over a 100MH channel on four by four MIMO, 256QAM, giving 1.3Gbps and 334Mbps DL/UL.

Chungwha Telecom

Another company with an AI-aided facial recognition app was Taiwan operator Chunghwa Telecom. The demo operator took a picture of your correspondent’s face, and saved it in a database. The little blue robot then filmed my face “live” and the edge-based AI platform instantly matched that to my face in the database.


The edge core

Aside from the “apps at the edge” use case of edge cloud infrastructure, another proposed driver has been the distribution of mobile core network functions to locations across the network. The aim here is to be able to provide core network support at locations closer to the network, perhaps in time extending to facilitating the concept of network slicing in a commercial way.


The ADVA Optical, Mavenir and BT demo that involved vEPC and vBBU elements on an edge server to support Network Slicing was also covered in our first article on the OpenRAN.  


Kontron was showing a vBBU integration from Altiostar on its ME1100 edge server. It also showed open source support in integrated hardware and software offerings for central office type applications, as dicsussed below.


Cloud infrastructure company QCT was showing its CORD-based “Central Office 2.0” solutions – including in a form factor for edge deployments. Here is one conceptual deployment in its lego city model. It calls its edge deployment of Intel-based NFVi a CORD-Ready POD. With Red Hat and OpenStack support it says this is the first fully-integrated open source infrastructure. Kontron (above) may have its own view on that – given it too has OpenStack-based and containerised deployment options.

Parallel Wireless

This company, perhaps better known for its low cost radio products, was showing implementation of analytics on its edge node – that it calls Het Net Gateway. The company was demonstrating the ability to containerise analytics pools as microservices on the virtualised infrastructure element. 

What have we missed? Let us know in the comments, send us an email.