The Edge of Confusion: Edge Data Centers
I hereby proclaim 2018 the year of the Edge data center.
I feel safe in making this proclamation since everyone seems to pretty much agree that, for a variety of reasons, data storage and processing capability needs to move as close to end users as possible.
I fear, however, that this may be the only thing we all agree on regarding edge facilities for the foreseeable future. This assertion by no means should be taken as reflecting my having any degree of pessimism regarding the opportunity presented by the shifting geographic needs for data center locations. It does, however, reflect the degree of confusion that already appears to be emerging around the actual definition of the term “edge data center”.
While seemingly a small matter, the lack of a universally agreed upon understanding of what defines an edge data center sets us upon a path of confused customers, conflicting analyst quadrants and some bad marketing. Have we forgotten the whole “modular” debacle which culminated in a variety of vendor assertions that roughly came down to “I’m more modular than you”? Do we really want to re-embark down the path of edge, edgier and edgiest? I think that prospect should produce a collective shudder in all of us.
What is an Edge Data Center?
Since it’s a data center, the best place to start is size. By implication, I think we can all agree that ”hyperscale” would not be the descriptive phrase one would expect to see listed in any edge facility propaganda. So how small do you have to be, and what is the proper unit of measurement, to qualify for the “edge category?” Based on their desired function, the number of racks is the most logical unit of size determination in the edge space. It is important to note, however, that a single rack offering is not an edge solution, but rather, would be considered a “micro” data center (defined as “data centers” of one rack and below). Naturally, there will be sub-classifications of “micro”—Lilliputian?—but that is a philosophical discussion for another blog.
Due to the expected dispersed nature of their implementation, edge data center power requirements are going to be a little less than their larger counterparts. Think kW, not MW.
“Inclusive” seems to be a popular societal requirement nowadays, and an edge data center should be no different. Certainly, it should include racks (refer to the size section above for reference), but also everything that we expect to find in their larger brethren should be part of an edge offering. Requiring end users to purchase certain components, rack, cable infrastructure, generators or chilled water for example, separately is an awful lot like saying “batteries not included,” and no one likes that.
I don’t think anyone will argue that “edge, » in data center parlance, is now a synonym for “remote. » And since one man’s remote is another man’s Cleveland, geographic independence is another basic edge data center requirement. For those of you saying, “Well, Wyoming is pretty remote,” I say, yes, it is but no one is filling a ski run at Jackson Hole with a 1,000 of these things, and nobody likes a smart-aleck.
Remote is often a term used to imply hard to get to, so the need for on-site personnel to perform operations and maintenance functionality would seem to be a disqualifying requirement for an edge data center. This isn’t to say that on the odd occasion someone needs to do some hands-on work at the site, but “lights out” operation should be the rule and not the exception here. These facilities shouldn’t require on-site operational personnel so if your new site necessitates that you to add a few more parking spaces for the engineers, an edge data center it isn’t.
An area for potential confusion is the whole colocation versus dedicated thing. What this really boils down to is the difference between an edge application and an edge data center. Yes, a colo facility in an office building in Boise can likely host edge applications from out-of-towners, but it’s not an edge data center. It’s kind of like saying all men are humans, but all humans are not men. In other words, it’s a distinction with a difference. So, in summary, edge data centers should have everything its big brother data center to standalone, run in a lights out fashion, and be measured in terms of racks and kW – not MW.
Often times technology brings confusion along with its benefits. Agreement on what constitutes an edge data center has the potential to be a long-term conundrum. I suppose it’s incumbent on all of us to do our best to keep the product of this next evolution of data centers from lapsing into ambiguity. Of course, I’ve seen some early market entrants describe their edge offerings as “modular” so maybe there are some issues that we’ll never solve.