Netflix’s VP of networks, Dave Temkin, poured cold water on the wave of edge computing enthusiasm when he said there was “no performance or cost benefit to be gained by a few milliseconds [of reduced latency]”. This would have come as a blow to many organizations which tout content delivery as the core use case for edge computing, especially when deployed by telcos.
However, what such comments should really do is push all stakeholders to be more precise, and more realistic, in their claims. Video on demand does not need edge computing, especially when companies like Netflix have done a lot of work to mitigate the effects of latency on their user experience. Live streamed video, or content based on virtual reality, would be a different story, especially for users on a wireless network. Other, industrial and IoT use cases will require even lower latency, and will rely on edge computing for other reasons too, notably security, privacy, local control of data, and reduced cost of transporting information to the cloud.
Most of these use cases have something in common with those for 5G, whose best attribute is also low latency – they are not fully proven as yet. The enterprise interest is real, but business cases – for those enterprises, and for the edge and connectivity providers, have not been fully made. So 2019 will not be the year of massive breakthrough for 5G or edge computing, but it will be the year when a great deal of work is done on solidifying the business models – and accepting that, by contrast with 4G or, indeed, the public cloud, those will be highly fragmented between different industry sectors and applications.
This is one of the challenges for mobile operators which want to achieve a significant place in the value chain in 5G or edge computing. They will need to address many different sets of requirements and groups of stakeholders, which will fragment their investments and their business model. That will be alien to a group which has basically pursued two models, voice and data, both with very broad applicability.
For the same reasons, many initiatives in edge computing will be driven by enterprise – and the same will be seen, to a lesser extent, in 5G, where many industries are pushing for private networks and even their own spectrum.
In the less developed market for edge computing, we are still at the stage where alliances and standards groups are behind many of the developments, but it is already notable that representatives of vertical industries outside telecoms and cloud are taking an active role. Success for operators, then, will be about absorbing the requirements of other sectors and not just understanding how to support them, but also engaging in active partnerships with them.
The risk that, rather than being trusted and essential partners, operators will be sidelined in the value chain, was seen at the turn of the year with the creation of a mega-alliance focused on edge computing, mainly for the IoT. This merges two highly influential groups, but both only marginally influenced by operators – the OpenFog Consortium (OFC) and the Industrial Internet Consortium (IIC). The combined organization will increase the momentum behind edge computing specifically in the IoT world, which is the one where operators say they have the biggest challenge to make a business case.
The OFC was set up to create an open platform based on the fog technology originally devised by Cisco. This defines a framework for filtering, processing and analyzing data at the edge, to reduce strain on the central cloud and associated connectivity. While the telecoms industry was focused on an alternative approach, ETSI’s Multi-access Edge Computing (MEC), OpenFog took a broader view, defining a platform which could be adapted for any industry, and which allowed the edge to be placed anywhere from near the data center to the ‘extreme edge’ or device. The broader applicability of this framework, compared to MEC, was seen when the IEEE adopted OpenFog as the basis of its 1934 edge computing standards, while MEC is now looking mainly at APIs (application programming interfaces) rather than a full architecture.
Meanwhile, the IIC was initiated by US industrial giant GE, which has also distanced itself from telcos in other aspects of its broad-based Industrial IoT strategy – including in its plans to acquire its own spectrum licences in the 3.5 GHz CBRS band. GE itself is going through a turbulent phase, but the IIC remains powerful, and has grown to be a broad industry alliance, very much led by the needs of manufacturing and other heavy industries. Telcos can take part, and AT&T is the leading participant from this sector, but their role is clearly confined to connectivity, rather than taking the anchor role in the value chain as they would have done in the MEC view of the world.
So now these two organizations are joining forces. They already have a crossover of major members, such as Cisco and Microsoft. “By expanding our pool of resources and expert collaborators, we will continue to accelerate the adoption of not only fog, but a wealth of technologies that provide the underpinnings to IoT, AI and 5G,” wrote Matt Vasey, chairman and president of the OFC, in a blog post about the merger.
Forthcoming work will focus on the increasing ability for machines and devices to have their own computing and storage resources – as smartphones, of course, already do – and to be somewhat autonomous and self-aware. “Robots, drones and self-driving cars are early indicators of small and mobile clouds. Distributed intelligence that interacts directly with the world and is immersive with all aspects of their surrounding is the concept behind fog,” wrote Vasey.
Although the groups each initially appeared to address a different sector and platform, their merger is logical given the inter-dependence of edge and the IoT, and the common challenges for both. The groups have also shared a similar overall approach, looking to reduce risk for enterprise adopters and so accelerate uptake, by developing common specifications and testbeds based on real world requirements. OpenFog has developed a full architecture, but not an academic or bluesky one – and the fact that this will underpin IEEE standards, making fog nodes and devices interoperable, is its most significant contribution to enabling a low risk, standards-based, mass-scale edge platform.
In 2019, the merged organization plans to increase investment in international testbeds and working groups, to remove its somewhat US-centric image, and to create new technical committees focused on horizontal features, but also on specific requirements for vertical markets. The combined consortium will continue with its core work of identifying prototype use cases in order to define best practice in terms of architecture for each sector. It also aims to add compliance and certification programs to its testbed activities, in which it will no doubt be helped by its close association with the IEEE, and its official standards body processes.
In the same week, another new edge group was announced, this one specific to Europe. The Edge Computing Consortium Europe (ECCE) consists, initially, of 18 vendors and other organizations, which plan to create a standard reference architecture and technology stack targeted at operators, smart factories, and other major Industrial IoT verticals.
The group said its first work would consist of four main areas of activity:
- Specifying the ‘reference architecture model for edge computing’ (ECCE RAMEC)
- Developing reference technology stacks called ECCE edge nodes
- Identifying gaps and recommending best practice within multiple scenarios (ECCE Pathfinders)
- Synchronizing with related initiatives and with standards bodies, and promoting results.
The founder members are Huawei, Analog Devices, ARM, Bombardier, B&R Automation, Harting IT, IBM, Intel, robotics firm Kuka, National Instruments, Renesas Electronics, Schneider Electric, Software AG, Spirent and TTTech; plus three research organizations from Germany – Fraunhofer Institute for Open Communication Systems (Fokus), German Edge Cloud and German Research Center for Artificial Intelligence (DFKI).
The group’s opening statements say that it predicts that 75% of enterprise generated data will be created and processed outside data centers or the cloud by 2025, up from less than 20% today.
Robotics firm Kuka said the edge is not just about processing data locally rather than sending it all to the data center. It is also about centralizing tasks that are currently done on devices, such as data filtering and pre-processing, to increase efficiency. It example was the common programming and configuration of several robots by an edge cloud controller.
Aerospace firm Bombardier said that edge computing would be a key technology underpinning cyber-physical systems, so the company was keen to drive appropriate standards that will support its global requirements.
“This new European edge initiative will fill the gap between existing industry initiatives and will enable German and European industrial partners to meet their business requirements enabled by edge computing in a faster and more economical way,” said Professor Thomas Magedanz, director of software-based networks for Fokus.
The groups which are defining open platforms for the cloud and the telco network, from OpenStack to the Open Networking Foundation (ONF), are increasingly focused on the edge too. At its recent ONF Connect meeting in Silicon Valley, the ONF ran a session on one of its ‘trailblazer’ work items, focused on how MEC would work with the ONF’s four reference designs – SEBA (SDN-enabled broadband access); NFV Fabric; UPAN (unified programmable and automated network); and ODTN (open disaggregated transport network).
“For almost a year we’ve been exploring with our partners, both our operator partners and our supply chain partners, what is the role that ONF can play in the context of this whole use case that is coming along that is called edge and edge cloud?” Guru Parulker, ONF’s executive director, told the conference. “The fact that we have built CORD and some of these reference designs we find ourselves in the situation that maybe ONF can play a role in enabling an edge cloud.” CORD (Central office re-architected as a datacentre) is the ONF’s foundational platform and underpins its vision of a virtualized, disaggregated network, as well as the four reference designs.
ONF’s CTO, Larry Peterson, added more detail about the intersection between the group’s work and MEC. “It’s really trying to put the puzzle pieces together and figure out how can we navigate this conversation, basically between the data centers moving to the edge and the edges adopting data center technologies,” he said. “These are two very strong bits of momentum and they are converging.”
As he pointed out, the edge involves both “access and software services in the same platform”, which has led to operators assuming, because they control much of the access, that they will also have the upper hand when monetizing edge services and applications. But as in any area where operators seek to move up the value chain from pure connectivity, they may not be best placed to take the biggest role.
But the ONF, of course, is telco-driven, unlike OpenFog and others, and so is an important theater of engagement for operators which believe the edge offers them an opportunity to get into cloud services. Many, especially on the mobile side, have surrendered in the effort to compete in the public or hybrid cloud market. But at the edge, the distributed and hyper-connected nature of the envisaged architecture may fit better with the operator’s processes and economics than with the highly cost-efficient, centralized and consolidated model of the traditional cloud.
So Peterson was keen to stress that edge platforms are not easy, and not necessarily easier for data center players than for operators. “It’s not just a matter of taking a large data center and shrinking it down into a rack or half a rack,” he said. “There’s really something fundamentally different that’s going on at the edge, at least when you start to bring access into it.”
To address that critical issue, the ONF has currently narrowed its edge focus to the ‘access edge’, and the work will be driven by its operator members, including AT&T, China Telecom, China Unicom, Comcast, Deutsche Telekom, NTT and Turk Telekom, as well as Google. The ONF defines the access edge as the location where virtual network functions (VNFs), localized evolved packet cores (EPCs), broadband network gateways (BNGs) and end services coexist.
“At ONF we’ve been asking, with our partners, two important questions,” Peterson went on in his keynote address. “One, is the edge, the access edge at least, just a smaller data center or is it something fundamentally different – qualitatively a different kind of platform?
“Two, what is it that we do in the ONF community, this operator-led community, that’s unique, that we can bring to bear on this problem? Clearly a it’s big space. We don’t want to try to do everything, but what can we uniquely do?”
Only after the ONF has made a significant contribution to defining the access edge and its architecture, should it move on to “figure out how to do end-to-end connectivity”, although all its work should be done with that platform, and what it calls the “service mesh”, in mind.
The motivation to concentrate on the access edge first is clear for a telco-driven organization – if the ONF can define the ground rules for this, its members will have an in-built advantage in driving a view of edge computing that aligns well to their own strengths and platforms. “In the access edge, that’s where you’ve got mobility supported,” Peterson added. “So the advantage that the telcos have is not just the points-of-presence near the end users. They also support ability. And 5G is embracing this wholeheartedly.”
But a key question – and one more easily addressed by MNOs than cloud providers or even telcos – is “how do you simultaneously move data to the end, to the edge, and allow for that functionality to be mobile? Simultaneously supporting those is really the hard problem” because the network functionality, as well as the connection itself, needs to be pushed close to the user, and in some cases, fully mobile.
While the ONF is clearly making a play for MNOs to lead the definition of the edge access platform, it also recognizes that the edge is not a market where one group of companies can rule alone. “This is not an ‘I own everything and I can solve the problem in a single thrust’ domain,” Peterson said. “We really have to deal with the different entities being able to contribute to, control, own, dictate activity at different points in this multi-edge environment.”
Other issues he highlighted were the need to ensure that solutions could be monetized and drive new revenues; that it must be easy to decide what is processed or stored at the edge, and what in the center, to achieve optimal cost efficiency; and
Peterson’s third point about the access edge was that ONF’s operator and vendor members need to deploy solutions that increase revenue streams. It seems obvious, but operators have historically often worked in a ‘build and they will come’ mode. But in edge computing and 5G alike, that will not work, at least not for the enterprise and IoT use cases which are expected to drive ROI. Those will require build-outs to be carefully targeted in terms of location and capability, and to be aligned to specific revenue opportunities – a cover-all network of dense 5G cells and edge compute nodes will clearly be logistically and financially impossible, so operators need to supplement their outdoor macro network and revenues with closely targeted build-outs.
Peterson told his audience: “Monetization is clearly going to dictate that you want to customize and differentiate services. And probably even as fine a grain as the subscriber level.” One important aspect of ensuring build-out costs align to revenue is to eliminate unnecessary expense by ensuring that some data and processing remains in the central data center. For different applications, there will be different balances of edge vs center, and of autonomy and intelligence at the edge. It will be important to success to figure out the optimal balance for each scenario – if operators can do this successfully, they will score important points over cloud providers, which face a similar difficult transition from rolling out a common infrastructure for anyone to use, towards more customized and specialized systems.
The ONF aims to ease the path for operators by capturing the requirements for some of the key scenarios. Comcast’s VP of access architecture, Rob Howald, endorsed the approach in his own conference keynote. He said: “Right while we’re in the middle of looking at what we should be doing for edge and access at Comcast, lo and behold, there’s an organization, this organization, that’s focused there as well. That part of its mission—focus on the edge and the access—may be synergistic with what we’re trying to do.”
However – as ETSI’s MEC group also acknowledged in its most recent white papers about edge business cases – the Foundation knows that its members will need to find their place in a complex value chain, not expect to have the whole edge market to themselves. The ONF is focusing on a multi-tenant environment that could include cloud providers, specialist edge cloud providers, over-the-top service providers and other third parties alongside telcos.” For instance, the ONF recently conducted a joint demonstration with Google, of CORD running with the Google Cloud platform.
“These are all players that we have to account for, and we’re just trying to not pick winners from a technology point of view at this point,” Peterson said. “We’re trying to keep the space open.”
SK Telecom extends edge cooperation with Deutsche Telekom:
SK Telecom has extended its cooperation with MobiledgeX, the edge computing initiative set up by Deutsche Telekom. The two organizations have signed a memorandum of understanding (MoU) to work together to develop edge computing technologies and business models.
Importantly, in a market where there are several non-interoperable platforms developing separately, the groups will integrate their respective edge computing platforms and work on a set of specifications which could be adopted more broadly to ensure interworking between edge networks deployed by different operators. This is expected to focus not on physical interworking, which should largely be addressed by IEEE 1934/OpenFog, but on APIs and higher level integration, to help expand the applications and service ecosystem.
This work will complement another partnership, between MobiledgeX, DT and Intel, to develop vendor neutral APIs and software tools for mobile-centric edge services, and make them widely available via the Facebook-led Telecom Infra Project.
“In the 5G era, mobile edge computing will be a key technology for next generation industries including realistic media and autonomous driving,” said Park Jong-kwan, head of SK Telecom’s network technology R&D Center.
In October, SK Telecom announced closer ties with DT and MobiledgeX, taking an investment of undisclosed size in the latter, while DT said it would inject the same sum into the Korean operator’s ID Quantique, a Switzerland-based quantum cryptography communication technology project.
That agreement built upon a strategic partnership signed at last year’s Mobile World Congress, when the telcos said they would share technologies and expertise, particularly in low latency services such as telemedicine; AR and VR; and security.
Tim Höttges, CEO of Deutsche Telekom, said in October: “We look forward to intensifying our successful cooperation with SK Telecom. The partnership will help both companies to strengthen our global technology leadership and bring 5G and other innovative services to our customers.”