Another year, another trend in the data center world. Although edge data centers first starting making headlines circa 2014 or 2015, they’ve become mainstream as more and more users slurp down increasing amounts of data. That takes serious bandwidth; to the point that many pundits are pointing towards the placement of workloads in edge facilities, rather than the traditional centralized data centers in major markets, as a sign that cloud computing is starting to wane.
On the contrary, edge data centers serve to supplement and improve the reach of even the major cloud computing providers. No major cloud service provider (CSP) is going to only place workloads in major markets. Just look at our neighbors in Cheyenne: Microsoft has a huge facility that they’re actively expanding. Amazon operates data centers in Ohio, which, while central for the US in general and equidistant from major population centers like Chicago and New York, is hardly a major market in itself.
And beyond large scale platforms like Azure or AWS, you have players like Green House Data, who offer smaller scale virtualization from data centers in a myriad of second and third tier markets.
But it's not just about the cloud spreading itself to the edge. Here's why edge computing will be important, but will also become more of a niche deployment model, with cloud remaining the king of application processing and data storage.
A sure sign that “edge” is reaching peak hype levels? Gartner has plopped it on its Peak of Inflated Expectations area of the infamous Hype Cycle. That means we aren’t far off from the big fall into the Trough of Disillusionment.
But why are data centers on the edge of the network such a big deal right now? A lot of it has to do with the growth of smart devices, the Internet of Things, and latency in general. Netflix or other data-heavy application performance can be improved if the information resides near the end user. And processing time for such (largely theoretical or in-testing) applications as self-driving cars can literally be the difference between life or death.
Meanwhile, some specific applications may require processing power far away from the prime network trunks that crisscross the country and terminate in major interconnection centers like the Westin Building Exchange; they may only have very limited bandwidth available, making long-distance processing and the return trip inviable. A remote industrial site, for example, or an armed forces unit deployed abroad.
These use cases are totally legitimate, but they aren’t that widespread just yet. Besides, many of the smart devices in use will live near a major interconnection point, and bandwidth is improving pretty much everywhere, all the time, with new submarine cables and wireless technology spreading higher connection speeds across the globe.
In most cases, as demonstrated by this article on Data Center Knowledge, cloud-native applications will splinter and deliver pieces of their service from different locations, rather than managing 50 or 100 deployments at edge sites across the planet. That’s simply too much overhead in administration and operational expenses, especially when you consider the security and monitoring on top of testing and deployment at each location.
Beyond the expansion of cloud platforms into edge data center facilities, other computing applications are likely to be deployed on specialized appliances — think about a piece of heavy machinery on a manufacturing floor, for example, that has built in processing features and must react in real-time. Meanwhile, the data collected by those appliances will still wind up in a cloud database, being processed and accessed via cloud applications, in a centralized facility.
Our own research in 2016 predicted these edge adoption trends. While interest in edge computing is strong, few had deployed anything in edge facilities and even fewer had plans to do so within the next year. That survey shows many of the perceived benefits of edge, but also some hesitation that seems to have continued into 2018.
Cloud computing emerged because of the efficiencies of scale. While bandwidth constraints may cut into those efficiencies for some specific applications, edge computing will not be killing off the cloud any time soon.