“You hear about the self-driving cars and the gaming and the VR. That’s all sizzle.”
Much buzz continues to surround edge computing. What does it really entail, and what are the benefits? At the Red Hat Summit in Boston, a panel of experts set out to define it and interrogate the issues with the technology right now.
Beth Cohen a software defined networking technologist at Verizon said: “So in my mind edge computing is taking the concept of the cloud in the data center, and blowing it out over a WAN network… A lot of people equate it with low latency, but it’s not necessarily just latency that is part of the definition.”
“I’d like to add that a lot of people associate edge computing with telecommunications and I think that’s a mistake. Because yes, obviously there’s a network involved, but I think the more exciting things happening at the edge are applications; so smart cities, IoT, and a whole bunch of stuff that is much more than just a 5G.”
Frank Zdarsky, a senior principal software engineer at Red Hat: “Over the last 10 years we’ve seen this trend towards centralisation; to very few, very large data centres. Edge computing for me is a counter trend to this. You have that because there new technical and business constraints at play that force you to provide certain services closer to the service consumer or to the data sources.”
What’s Driving the Uptake?
Lower latency is one of the key driving factors in the uptake of edge computing technology, as many enterprises see the advantage for their clients in having a service that can respond as quickly and efficiently as possible.
The efficient utilisation of bandwidth is also often cited as a driving force for edge computing technology.
Take an organisation with a thousand security cameras feeding back HD streams to a storage system: it’s not efficient to just record everything; instead, edge computing can give those cameras the intelligence to know what is important, for instance when the sensors detect movement or a change in lighting, handling a large degree of compute at the edge and minimising the load sent to the cloud for further analysis.
Steve Vandris, Intel’s edge computing infrastructure lead, noted that privacy is another factor in the rise of the edge: “[In Europe there’s much more emphasis on] data privacy and data sovereignty. So many enterprises – either for regulation or for proprietary reasons – don’t want their data to exit the corporation… many large industrial corporations don’t want their data to go outside: whether banks, or healthcare, there’s huge public relations issues on privacy, so that’s another additional big driver.”
Environmental, Geographical Issues
Daniel Bernier, a networking strategist with Bell Canada, Canada’s largest telecommunications company, noted that one problem for his team is that edge devices by their nature are far from the technicians and in a country as large as Canada it can be a real issue. Some of the device may only be visited every six months. One of his engineers was recently delayed by a passing herd of Caribou, he joked, adding that effective edge computing in large areas can entail major costs.
“[There are cost concerns] while we have uncertainty over the actual applications and the workloads which are going to be released [to run on the edge]. We’re missing the the ‘slam dunk use case’ we call it internally, that makes it justifiable to build everything [at the edge] for us. We have a big country… we have like things up north near Santa, so it’s really complex.”
Orchestrating the Edge…
A second problem for him is how to manage massive orchestration, as edge computing is going to comprise – in some cases – thousands of different distributed devices.
“[A challenge is being able to] move workloads and figure out the best placements of workloads. There’s actually a lot of work on this with using machine learning and AI to be able to figure out that model of how to place workloads efficiently”
Cohen states that stability and connectivity is still a big concern with the deployment of edge computing: “It requires a high degree of automation over WAN networks and over unreliable networks, so unlike a data centre where you can write a lot of automation but you’re also going to assume that that top rack switch is going to run pretty much 100 percent of the time, you cannot guarantee that in an edge environment.”
This also leads to the issue of physical security. Data centres are secured environments. With edge computing your hardware is out in the open, and in the example of Bell Canada, may actually be a helicopter ride away, so ensuring that devices out in the open are not tampered with can start to become a problem.
As Cohen concluded: “Too many people are thinking of the edge as a telco play; you know if you read all the press headlines it’s 5G/edge. Don’t think of it that way. There’s a whole bunch of other applications out there that are driving the edge as well.”
“You hear about the self-driving cars and the gaming and the VR and that’s all sizzle. I’m sure it’ll come eventually, but what’s going to drive edge computing is business cases that financial sense. Why are we using edge for the telecoms thing? It makes financial sense. Why are edge IoT gateways showing up? Because it makes financial sense, and that’s coming out of the enterprise and it’s coming out of telecoms.”