The network edge used to be fixed and predictable in terms of location and capacity needs. But these days it is neither – the edge has become a moving target, which is creating challenges for cloud players and data center operators trying to provide the capacity necessary to support edge customers.
That was the overall theme from a roundtable session at PTC 2017 last week that focused on the strategic importance of the edge as mobile apps, the cloud and the IoT increasingly push content to the network edge (and thus to the user). This has resulted in a complex series of business and legal relationships between various players, from content providers, data center operators and telcos to internet exchange operators, content delivery networks and channel partners.
It’s also resulted in making the very notion of the edge more nebulous, said Phill Lawson-Shanks, chief architect and VP of Innovation at EdgeconneX.
“The definition of the edge is evolving. Seven or eight years ago, you’d have said that the edge is where the infrastructure meets the towers,” he said. “Later, you’d have said it’s where the caching nodes meet the end points. But now it’s more about defining it as the critical infrastructure where you can get the content and transaction processing at the closest possible point to the end users.”
Andrew Schaap, senior sales VP at Digital Realty, added that as the IoT starts to come into play, “the edge is becoming closer to everything for different users.”
Not only that, said Frank Rey, director of Microsoft’s Global Network Acquisition Group, the edge has actually become mobile.
“The edge has traditionally been a fixed point, but now it’s becoming a moving target as we connect more things that move around – cars, cows, whatever,” he said. “That will affect how people set up edge networks. It’s no longer about just having one or two sites in a given market.”
Lawson-Shanks said this was why EdgeconneX’s business model is to work only with anchor tenants. “It’s not build-it-and-they-will-come, unlike the traditional data center market. It’s a partnership. So for example, the cable TV companies will say they have to have their equipment infrastructure in this market, so we’ll look at how their infrastructure sits and do something that will be close enough to their on-ramps – headends or COs or whatever they have – so that it’s passive internet access. Then we’ll build the ecosystem. But it’s very much customer-driven.”
Schaap added that customer behavior is also a factor because different verticals have different needs. “A hospital doesn’t have the same requirements as an airline. Or another customer says they need to be close to Azure. This also means that it’s no longer driven by geography in the sense that people aren’t focused on locations like New York or Chicago.”
Rey of Microsoft noted that geography is still somewhat relevant in terms of things like population size: “You have lots of people in New York or London, for example.” But he did agree that “it’s still what they’re doing that will dictate what your network looks like.”
The human behavior problem
One particular problem for edge players is knowing how much capacity a given data center customer will need. This usually involves forecast models based on customer requirements. The problem is that many customers aren’t really sure what they need, which can throw forecasting models off.
A key reason for this, said Rey, is many companies have not yet moved their data to the cloud, which brings in an unpredictable human-behavior element into the mix.
“We get situations like this one time a utility company wanted to close their data centers, and they decided to move all of the workloads of one data center to Azure as a test,” he explained. “When they saw that it worked well, they decided to move all eight data centers to Azure – not phased over a couple of years but all of them, right now, and they wanted this much capacity to do it. You can’t predict something like that – it’s not that our forecast models are incorrect, it’s that our customers are going through their own transitions, and that human behavior is unpredictable. Sure, it’s better to have that problem than the reverse, but it’s still a challenge.”
Schaap of Digital Realty agreed. “There are CIOs out there who have been told – probably by board members – to move to the cloud, but they’ve got 2,700 applications and they have no idea what it’s going to look like when they move parts of it to Azure and parts to this ERP manufacturer and parts to this other provider. So they have a hard time getting a prediction. When you look at the supply chain, [Microsoft] is changing that right down to the chip, so it’s hard for enterprises to get their arms around it and explain to suppliers what they think they want.”
He added that there are ways to help give them line of sight further down the line and future-proof for them, “but human behavior is hard to predict.”