Over the last two decades, hyperscale datacenters have reshaped the computing industry, making possible cloud computing, online retail, and media streaming on vast scales. But today’s centralized model is starting to yield to a much more distributed infrastructure that is being driven by a set of applications that must consume and process data where is generated.
What we are talking about, of course, is edge computing, a new model that is being defined by applications related to IoT, 5G, augmented reality, online gaming, smart factories, and autonomous vehicles. In these environments, the requirements for low latency and high bandwidth is pushing the infrastructure a lot close to the source of the data, whether it’s stored or streaming.
That doesn’t mean datacenters are going away, but it does mean there will be a lot more of them and some will barely resemble datacenters at all. A company that jumped into this space early is EdgeConneX, which builds a wide array of edge facilities for different use cases. These centers can range from tens of kilowatts up to tens of megawatts. Today the company has an installed base of 200 megawatts, spread across 36 facilities in North America, Europe, and Latin America.
EdgeConneX’s customers mostly consist of some of the largest IT providers on the planet, including those in cloud services, media content distribution, IoT, and networking/telecommunication. The company is strictly focused on designing, constructing, and operating facilities; the actual computing infrastructure is supplied by their customers.
The big challenges for them are securing real estate, power and cooling, often in some rather datacenter-unfriendly locales. Often these facilities have to be built in the middle of cities, where vacant land is rare, extra electrical grid capacity is limited, and there is barely any space for cooling units. In some cases, they have deployed on-site power generators to supply the centers.
According to Matt Larbey, EdgeConneX’s Managing Director for Europe, they’ve gotten really good at locating real estate, power, and fibre, and are able to build datacenters quickly in the locations their customers need it.
EdgeConneX’s business model is entirely demand-focused. They will deploy a datacenter only after they have found an anchor customer for a particular location. “We don’t turn up in a market, build a datacenter, and say ‘hey come to me, I’m now an operator in this market,’” explained Larbey. “We only ever build where there is a lead use case for that datacenter.”
The company has been around since 2009, but didn’t really get into the datacenter business until 2013. Prior to that they were providing a micro-edge solution, known as EdgePop, for telecommunication companies. Essentially this was a 30 kilowatt computer cabinet that sat at the base of a cell tower.
After 2013, they got noticed by network providers that wanted something larger than an EdgePop, to cache media content for local customers. At the time, networks were getting hammered with the deluge of data for things like YouTube and Netflix, which was all being funneled through just six main network exchanges in the United States.
EdgeConneX ended up building about two dozen additional network interchange points in places like Minneapolis, Madison Tallahassee, and Nashville – cities without existing quality datacenter capacity. These edge centers were about 1 megawatts to 4 megawatts in capacity.
“So content really drove what we term Edge 1.0, which was content caching and localization,” Larbey told us.
In 2014 and 2015, EdgeConneX ended up on the radar of the big hyperscale operators and software-as-a-service providers who needed cloud infrastructure closer to their consumers for latency and bandwidth sensitive services. The datacenters acted as local access nodes for things like CRM, cloud computing and storage, as well as for providing an on-ramp to the hyperscale centers. Since the use cases were varied and demanding, these datacenters were a good deal larger, reaching 4 megawatts to 8 megawatts, even 20 megawatts to 25 megawatts for the “edge core” centers.
But the bigger opportunity lies ahead, in what Larbey calls Edge 3.0, encompassing the emerging application areas we mentioned at the beginning of this article – IoT, augmented reality, autonomous vehicles and so on. Even though these would be primary served by micro-edge datacenters – in many cases just containers of servers and storage – there will be many more of them since the latency requirements are so stringent.
For these Edge 3.0 applications, Larbey foresees that each market, will get 5, 10, or 15 microedge centers, which is sure to draw a lot more competition. In fact, he’s already seeing multi-tenant datacenter providers and other companies pursing the micro-edge space.
“I think that shows you how buoyant that part of the market is,” said Larbey. “We are seeing significant growth and demand in that area. There is literally a tsunami of data that will need to be managed and the current infrastructure just can’t manage it.”