Operators try to keep control of edge compute momentum ahead of 5G

Deutsche Telekom is to place its edge compute activities in their own unit, to be branded MobiledgeX. This is just the latest announcement from an operator which is hoping to ride on the tide of interest in edge computing to support new revenues and business relationships.

But as we have argued before, this can be a double-edged sword for operators. While ETSI’s Multi-access Edge Computing (MEC) architecture has strong telco roots (and was originally mobile-specific), other edge compute platforms clearly envisage enterprises or cloud providers having the upper hand in the value chain. The rise of the connected edge will support the distribution of cloud services and content closer to the user. This will be important to reduce strain on the network when it comes to high bandwidth video or virtual reality content. And it will be essential to enable the ultra-low latency services which should be a central element of the 5G model.

All this will increase operating efficiencies for mobile operators, and help them to get more out of their 5G deployments with critical, ultra-low latency or massive IoT services. But will they seize a premium place in the value chain, or merely be the bit-pipes which link devices to edge gateways or local servers? In the latter scenario, they risk making less revenue from their core activity, carrying data traffic, since that traffic will be less frequently travelling long distances to the central core and cloud. And an even greater risk is that third parties will build out a lot of the edge infrastructure, taking advantage of shared spectrum and commodity hardware with virtualized functions. That could see a specialized ‘heavy MVNO’ – one which builds its own hardware in certain areas – creaming off a great deal of the value from an enterprise or IoT service, while targeted applications providers take much of the rest.

To avert such a poor return on 5G investments, operators need to work out now how they can inflate their role in mobile cloud services and especially those distributed to the edge.

Deutsche Telekom’s effort, according to reports in LightReading, will be led by Jason Hoffmann, who used to head up cloud activities at Ericsson and was also a founder of cloud computing start-up Joyent, now part of Samsung. The telco told the online journal that the aim is “to further explore the opportunities of edge computing” but was short on further details.

Other operators are getting active at the edge. We can expect to see deployments in venues at the Winter Olympics in South Korea next month, and the Summer Olympics in 2020 in Japan – stadium applications are an obvious low hanging fruit for mobile edge platforms.

Also, AT&T is planning to open a MEC innovation centre in California during this quarter. Last fall, the US operator provided an outline of its particular strategy for edge computing. “Edge compute is the next step in getting more out of our network, and we are busy putting together an edge computing architecture,” said Alicia Abella, a senior executive at AT&T Labs, in a keynote at the Fog World Congress.

There are many religious arguments about the location of the network edge, and where edge-based computing, storage and other distributed cloud resources should be located. Some telcos, including AT&T, are shifting from early enthusiasm for MEC, which integrates the processors tightly with the mobile network layout and places them as near the end user as possible, to more IT-driven architectures like OpenFog. This looks to place the resources anywhere in the network, from data center to user device, depending on use case, with the overall objective being to maximise resource efficiency and reduce latency, by sending less traffic to the central servers, easing the strain on the network itself.

“We want to deploy edge compute nodes in mobile data centers, in buildings, at customers’ locations and in our central offices. Where it is…depends on where there is demand, where we have spectrum, we are developing methods for optimizing the locations,” Abella explained.

Abella does not yet foresee very miniaturized edge systems, although AT&T has conducted demos running video surveillance applications on a Raspberry Pi. Edge systems “could be as small as that, but more realistically we foresee some as small as a walk-in closet,” she commented. She added: “Imagine a new class of devices, or AR/VR without tethering to a computer but using mobile-edge compute…imagine a path to access signal processing in the edge.”

AT&T aims to define an open software platform for edge computing so developers can write applications for it from 2020 – initially for its own use, but potentially later as a de facto standard. Abella believes that, eventually, such a platform could support crowdsourcing of processing, storage and connectivity in a way that would shake up the operator and enterprise cost base yet again.

Operators which have significant infrastructure and wireline assets, such as DT, AT&T and also the UK’s BT, have multiple ways to attack a distributed, edge-based world and monetize it. They have talked about reusing locations such as switching centers to house new cloud infrastructure, and could offer distributed infrastructure on a wholesale basis as well as to support their own retail services. The tower operators are thinking the same way – Crown Castle, for instance, plans to launch edge-based infrastructure-as-a-service this year, including IT resources along with its mobile cell sites.

All this will only be profitable if there is the predicted rise in usage, by consumers, enterprises and IoT devices, of services that will benefit from nearby connectivity, storage and processing resources. Most of the early attention has been on improving video quality of service, while reducing strain on the network, but in 5G, the low latency issues are likely to become more important if cellular networks really are to be used for critical communications, autonomous driving and drone management (though many telcos, including DT and BT, are still expressing doubts about these as commercial opportunities, at least in the near to medium term).

Consumer and industry uptake will rely on gadgets that can take advantage of all the nearby resources, supporting high end video formats, virtual reality and AI-enabled services. So it was no surprise that, at last week’s Consumer Electronics Show (CES) in Las Vegas, the chip suppliers were heavily focused on moving complex processing to the edge.

NXP demonstrated edge cloud applications based on its cooperations with Google Cloud Platform,  Amazon AWS, Microsoft Azure and Alibaba. They showed data-intensive applications such as facial recognition, as well as IoT sensors connecting to an access point close by for near-real time processing. NXP was providing the connectivity chips that link the local and centralized platforms, and the end user sensors or devices.

“We used to call things embedded computers,” said Sam Fuller, head of system solutions for NXP’s digital networking group, in an interview. “It’s not the first time that we’ve had computers that weren’t in the cloud, but what we really are creating here is that ability to create a framework whereby it’s very easy to develop software that collaboratively works between these edge devices and cloud devices.”

He also pointed out that, with edge computing, devices can access new features and skills without the hardware having to be upgraded each time. “What has really changed is that historically devices like that have been hard-wired. They’re really appliances, and they’re not platforms. And I think what edge computing does… [is] extend a programming model such that new types of applications can be developed that have a computing component that’s local. Might be in your set-top box. Might be in your home gateway. Might be at the base of a cell phone tower,” he told LightReading.

MediaTek was also focusing on the edge, and particularly on being an “edge AI enabler”, as CFO David Ku put it to journalists in Las Vegas. He said MediaTek was working on AI functionality that requires so little computational power that it could be deployed in mass devices like light switches.

Like most chip suppliers at CES, MediaTek was excited about voice activation and digital assistants – certainly the device category of the year in Las Vegas 2018. While Amazon is focusing on cloud-to-cloud services for its Echo devices, MediaTek is more interested in a hybrid model of edge and cloud, with more local intelligence, so that the smart speakers can distinguish human and non-human voices, cancel music in the background, and recognize vocal patterns.

MediaTek’s interest in AI-enabled light switches is related to this, since the most common use of smart speakers is supposedly telling lights to turn on and off. Ku said: “Why not add that small AI capability to light switches?” There would be no need for a voice assistant in each room,  just a small DSP and neural network in the switch, which would only need to recognize 20 to 30 key words.

This could be enabled by technology from a spin-out from MediaTek, Intelligo, in which the larger firm is still an investor. Intelligo’s Debussy SoC is based on an AI processor called iGo and is targeted as an intelligent voice processor which could, in time, be shrunk down to tiny devices.

All this is part of an urgent push by MediaTek to reduce reliance on the smartphone market, where the Taiwanese company has been losing share to Qualcomm and others, and to lead in certain IoT-related growth areas. Last year, the firm got about 30% of its sales from IoT and ASIC/PMIC units, and revenue from these areas rose by 20% year-on-year, with WiFi connectivity and voice assistant technology particularly strong. Another 30% came from the “stable cash cow business” – chips for digital TV, featurephones, monitors and optical storage.

Other newer departures are to provide ASICs for data center switches, which will take MediaTek up against Broadcom, among others; and four lines of automotive products – telematic modules, mmWave radars, in-vehicle infotainment and digital clusters.

Saguna pushes MEC forward again with Starter Kit:

Saguna is one of the earliest proponents of MEC – even before it was called MEC. The firm worked with Nokia and Intel five years ago on their Liquid Apps platform, which was essentially based around the same concepts which were embraced by the ETSI MEC initiative, in which Saguna (and Nokia) were also early leaders.

Last week, Saguna released its Starter Kit for MEC, which is said is designed to help operators and developers “gain valuable hands-on experience, realize the benefits and accelerate development, integration and deployment of their MEC ecosystem and edge applications”. The kit includes a complete, virtualized MEC environment, sample edge applications and support package.

In June, it was notable that Saguna and Nokia were once more taking an early leap into a potentially significant development for MEC – Amazon’s launch of its Greengrass edge cloud development platform. A deep alliance with Greengrass, and therefore with Amazon’s huge ecosystem, could give new impetus to the Nokia-led MEC world view, which places the MNO firmly at the center of the edge compute chain, rather than the IT-driven view, in which the mobile network is merely a pipe.

Saguna’s Open-RAN gateway, a MEC server which is deployed right inside the RAN, now supports Greengrass. CEO Lior Fite said: “We are proud that AWS Greengrass is now available on the mobile network edge using our MEC platform. This seamless meeting of cloud computing and communication networks creates new monetization opportunities for mobile operators and customers.”