MEC is delivering low-latency, real-time enterprise applications
As has been pointed out by operators on recent earnings calls, the market for advanced 5G-enabled services—things like mobile edge computing (MEC) combined with public and private 5G, and private 5G itself—has been slower to create revenues than previously thought. Despite that, the vision is unchanged the machine is in motion. Operators still see themselves, and partners ranging from hyperscalers and systems integrators to independent software vendors (ISVs) and application developers, as a lynchpin in delivering digital transformation, driving industrial efficiency, and the economic gains that come with that.
To understand market developments, including what enterprises want, how they want to consume it, and how relevant vendors are forging partnerships to bring MEC to market, we convened edge specialists at the recent Telco Cloud & Edge Forum (available on demand here) in an effort to make sense of it all. KORE Wireless Chief Information Officer Chris Francosky and Volt Active Data Chief Product Officer Dheeraj Remella joined RCR Wireless News to share their perspective on where MEC is today, where it’s going and why.
From Remella’s point of view, which aligns with Volt Active Data’s speciality, enterprises are trying to solve data problems. To do this, he said the company has focused in on four key capabilities needed to inject intelligent actions into streaming data; those are: the ability to store data, aggregate data, apply business rules to that data, then take appropriate action. “Typically there are like a plethora of technologies that would accomplish each of these capabilities…So when you’re looking for lower latency, having several technologies really becomes unmanageable or unscalable, and beats the fundamental of the objective, of the latency.”
To meet the latency objective and intelligent parse real data as it’s created, Volt Active Data has joined up with operators, hyperscalers and enterprises to monetize significant data events, identify and prevent threats for revenue assurance, and add operational efficiency with process automation. “These are the three broad categories where you would see really low latency, complex decisions, and several sources of data. “And bringing this all together at the edge becomes quintessential going forward,” Remella said.
For KORE, which has evolved from a sort of MVNO for internet of things (IoT) connectivity into a more turnkey solution provider, Francosky said the company is delivering more self-service, more automation tooling, more connectivity, more data exposure, and, in general, more scale to the world of IoT. “These devices are becoming more sophisticated, more capabilities are being added,” he said. “So it’s important that we have more distributed presence around the globe to support billions of devices…We’ve historically been focused on more just horizontal solutions, but recently have really seen a need to create these dedicated business units where we’re offering bundled solutions for fleet, healthcare, industrial IoT, more specialized solutions in those verticals.” More later on vendors moving from a horizontal to vertical approach.”
For enterprise digital transformation, does MEC solve new problems or solve old problems faster?
The answer to the above question is “both” really. More digital operations creating data means there are new problems around turning that data into operational improvements. And MEC, particularly alongside public and private 5G, can also take existing problems and solve them more quickly, more responsively.
Francosky hit on latency, the need for localized data interactions, better application responsiveness and data sovereignty as issues MEC can address. He also noted growing traction around combining cameras-as-a-sensor with intelligent video analytics platforms. “It’s nearly impossible to get near-real-time video analytics by calling an API that’s hosted 70 milliseconds away in some sort of cloud data center. So moving that closer to the actual device itself is critical for this.”
Remella recalled earlier IoT days when the focus was telemetry-based use cases wherein sensor data was used for various control systems. “But more recently…there’s a lot of video coming into play.” This is, in some ways, an extension of telemetry use cases but drawing in data sources that where previously set up in a way where pulling device telemetry wasn’t possible—“So what you are observing visually, you convert that into data and use that for making decisions.”
He gave the example of autonomous guided vehicles (AGVs) vs. autonomous mobile robots (AMRs). The former follows a pre-planned path and stops if there’s an obstacle and waits for it to move before continuing along that path. AMRs, on the other hand, “are capable of navigating around obstacles” by using computer vision. There’s a downside here though: “These AMRs become really, really expensive because there is so much intelligence built into it.” But with the ongoing buildout of edge, devices like AMRs, and others, can be commoditized as intelligence is taken off the device and put into a local cloud.
Big picture, Francosky said, enterprises are embracing AI and machine intelligence; in that continuum, video is just another type of data to be exploited. “Now that enterprises have kind of awoken, so to speak, knowing the data is the fuel—whether it’s textual data, or metadata, or log data, or even video or photos—more and more of this data needs to be collected in order to feed these models. As this trend continues, and as the proliferation of 5G speeds up time to compute, and localized compute speeds up time to action, “That’s where I see the real need to…minimize the path from the device back to that compute power,” he said.
What goes where? Where which data is processed—a MEC node or a central cloud—depends on the use case
Whether an enterprise is trying to implement computer vision or ensure data security/sovereignty, the first question enterprises need to ask themselves is which parts of their processes would improve with more responsive computing, Remella said. This will then prompt a further analysis of what data needs to processed locally and what can be processed away from the edge in a more traditional manner. “Cloud is not going away,” he said. “The cloud is necessary for the virtually unlimited resources that you get.” Splitting intelligence across these two pools of resources, “You’re making continuously evolving, better decisions. And that’s what we are actually seeing.”
The long goal, Remella said, is to facilitate development of “active digital twins” where a more classic digital twin is imbued with real-time context and intelligence. “It’s almost like an intelligence pipeline,” he said. “As far as well can tell, this is going to become mainstream…It’s an evolving thing.”
Francoskytracked a decade-plus shift from on-prem data centers to massively-scaled public clouds and now back to the edge with the maturation of 4G and now 5G. With cellular connectivity being what it is, and with the enterprise need for faster, more intelligent data-driven action, “It just doesn’t make sense to push that data back to a centralized cloud provider,” he said. “It’s happening gradually, but there [are] some killer use cases coming out that I think are going to accelerate this move to more of the edge and then, distributing computing in general.”
Examining this same shift in where compute takes place, Remella said the move away from centralized clouds and essentially open-ended access to computing resources to the edge, where computing is certainly not open-ended, requires a more nuanced approach. “Your resources are constrained,” he said. “When you actually bring [computing] to the edge and you need to make your decisions based on data, and both your data and your decisions need to consistent to reality, and it needs to happen with low latency…you cannot have so many layers because any of them could actually fail.” This, he said, points back to the need for a unified data management solution as opposed to a more piece-meal approach.
Should vendors become vertical experts, or should verticals become technologists?
Back to Francosky’s earlier comment on moving from building a horizontal platform to developing more vertically-focused solutions; he gave the example of KORE’s acquisition of connected healthcare specialist Integron and Twilio’s cellular IoT business.
“What we’ve done,” he said, “is we’ve taken some of our key verticals in addition to healthcare, fleet, industrial, for instance, and set up, let’s say, specific pods or business units that are really focused. With experts in those areas that understand the use cases, that understand the ecosystem, everything involved in those particular verticals. To just provide a, let’s say, more specialized expertise within those verticals.”
Given the variability in MEC use cases and attendant applications, he said, “I think where you’re going to see more of the processing investment in the actual compute and storage at the edge is going to be where you need…that real-time decision making.”
But, bottomline, Francosky said, “You’ve got this new paradigm. It’s a new architecture…The cloud, the advantage is it’s offering infinite scalability, [and] developers have taken advantage of that, whereas maybe at the edge application, they have limited resources to work with. So that’s new…The tools are evolving…These new tools are coming out to make it easier for developers to move from, say a cloud mindset out to a more distributed edge computing architectures.”
