Two weeks ago, we covered the IEEE’s decision to adopt OpenFog as the basis of its own IEEE 1934 standard for edge compute architectures. This was confirmation of something Wireless Watch has suspected for some time – that ETSI’s Multi-access Edge Computing (MEC) specifications are being relegated to a supporting role in the edge environment, and with them, the operators’ role in that value chain.
ETSI’s latest MEC white paper does nothing to dispel this idea. It outlines how MEC and 5G can work together to enhance the overall opportunity for the MNO, but it also focuses on how this combination can support other players in the value chain, including the cloud providers and Industrial IoT specialists. This is a necessary step, moving MEC away from a vision in which the edge was built around telcos’ locations, and the operator was the lynchpin. Instead, it recognizes that, while some telcos may become significant edge cloud service providers, most will have to squeeze as much of the total market value out of their connectivity as they can and work with other, more powerful edge players.
And while this makes MEC useful to a wider range of stakeholders – and the close relationship to the 3GPP and 5G will certainly be helpful – ETSI has missed the boat in terms of setting the leading direction and framework for edge compute. Just as in 5G itself, operators will now have to be skilful in making alliances and designing value-added services, to take a strong role in the vertical industries where most edge/5G growth will be found.
ETSI MEC originally set out a telco-centric vision of edge compute in which the edge nodes (localized processing and storage to support distributed cloud) would align neatly with telco assets such as central offices (COs) or cell sites. However, as operators complained that even macro sites were too dispersed an edge for most applications, other groups were defining the edge as being within enterprise premises (such as a factory floor) or even a car or home.
In this very dispersed cloud (mist, as Cisco calls it), there is no logical alignment of telco and edge locations. The operator will provide connectivity, of course, but the dominant players in the value chain will be those supporting cloud services for particular markets and the edge resources will be placed wherever those vertical markets require them. This is why OpenFog is more powerful than MEC, because it allows for nodes to be located anywhere from the central data center to the consumer, within a common framework. It will work with third party APIs, including ETSI’s, but it sets out an umbrella that can be harnessed for any value chain.
One card ETSI MEC could have played would be to converge its architecture with small cells and ride on the wave of densification which will occur with 5G.
Some operators will do this individually, supported by groups like Small Cell Forum, and that provides a neat combination of connectivity and cloud resource, as previewed years ago by Intel with a small cell companion processor, and by Nokia with its Liquid Apps system. Such options will continue to be developed, particularly to support video caching and content delivery networks (CDNs) – the areas on which most MNOs have focused early edge models.
But only a few MNOs are likely to become significant edge cloud providers across a wide range of verticals, and pushing themselves up the services stack. Instead, their success in enhancing their revenues and differentiation by exploiting edge compute will rely on partnerships with the real edge winners – and those look set to be a mixture of the webscale players and vertical industry specialist providers.
This was clearly reflected in the white paper, which was published earlier this month. ETSI MEC has already defocused on its reference architecture in favor of pushing APIs which can be used with less prescriptive frameworks. Now it is looking to make operators into attractive partners for industrial IoT and cloud providers by emphasizing the synergies of 5G with edge compute.
But reading between the lines of the paper, there is clear acceptance that the operator will be just one player in a chain, and not the central hub, as ETSI members once hoped they could be if they tied cell sites and edge nodes tightly together.
In the white paper, the group outlines its view of the role of MEC in 5G, and how edge cloud applications can be hosted in the 5G network.
As such, it covers MEC for:
• third party cloud service providers
• massive IoT – especially serverless computing and cloud integration of huge numbers of IoT devices
• enterprise users
• and Industrial IoT.
The argument is that 5G will
• increase the value of MEC to all these players because it includes significant edge compute enablers with which they can enhance their services
• and that, therefore, the cellular ecosystem will benefit from the MEC APIs to help enable applications and service platforms “at the very edge of mobile networks”.
The first is true, since 5G’s distributed cloud capabilities will support some innovative use cases and will help set the scene for programmable, sliceable networks that use edge resources in the most dynamic way. It is unclear, however, how much value MNOs will derive from that if they are playing a bitpipe role, or if vertical industry players are using shared spectrum to support 5G connectivity.
For the same reasons outlined above, that makes the second argument weaker, especially if the cellular industry continues to think of the “very edge” as being a few central offices supporting a few CDNs, rather than a platform to support applications and analytics everywhere that businesses need to collect data.
The paper talks about the ability for MEC-enabled applications to gain contextual information and real time awareness of their local environment through the standardized APIs. Some operators, and many web players, are designing services that harness just such capabilities. Some 5G frameworks are also focused on this concept, such as the UK 5G Innovation Center’s Flat Distributed Cloud architecture, which has advanced ideas about edge compute and hyper-contextual analysis. But such approaches will require edge nodes attached to enterprise buildings or small cells, not COs or even macro sites. And OpenFog, with its more flexible approach to the location of the edge, will get to that first. Its users may well harness MEC APIs to gain optimal 5G benefits, but that gives ETSI MEC, and the operators which rely on it for their edge cloud strategy, a minor role.
The white paper insists that the alignment of 3GPP standards and its own APIs and platforms still means there will be an enhanced role for MEC, and operators, in the 5G edge value chain. “The 5G system specification and its Service Based Architecture (SBA) leverage the service-based interactions between different network functions, aligning system operations with the network virtualization and software-defined networking paradigms,” says the document. “These very same characteristics are shared by ETSI MEC specifications.”
That angle just harks back to the idea that, because telcos are interested in edge, the edge will necessarily be fully aligned to their approaches – just when other industries are, for the first time, seeing the opportunity to seize 5G and turn it to their requirements, rather than just riding on a network designed around MNO priorities. These industries, supported by Industrial IoT platforms and by cloud giants, will be able to combine OpenFog (or other standards) with MEC APIs so they can best exploit the edge/5G combination, but still leave the MNOs down at the bottom of the service stack.
All this makes the white paper’s sections about third party cloud providers particularly interesting. This chapter begins by saying “MEC services are typically envisaged as being offered and supplied by Mobile Network Operators. However, a MEC service can also be offered by the third parties” – which, given the heritage outlined above, is a significant admission.
In providing more detail, the paper hits on the real threat to its MNO sponsors, saying: “For instance, third party cloud service providers are entities offering MEC application hosting services and resources, while not being traditional network operators. Examples of such third party providers willing to deploy edge cloud resources include: venue and facility owners or management companies, cell tower owners and neutral host vendors and vehicle fleet management companies (railway, automotive etc). Due to operational complexity, costs, or simply difficulty in deploying MEC in hard-to-reach areas (for example, due to real estate scarcity), MNOs may buy edge cloud services from such third party providers to supplement their networks. Typically, MNOs are not involved in day to day management and operation of these third party edge clouds.”
This acknowledges what many MNOs find hard to do – that in a dense network of small cell sites and edge nodes, as envisaged for 5G, they will have to accept a level of equipment sharing that has previously been unacceptable, if they are to make the economics work. It will make more economic sense to sit on edge nodes deployed by cloud companies and enterprises (though ETSI carefully does not mention the webscalers). And that means striking new balances of power and profit with partners, rather than controlling everything from their high cell towers.