This article is more than 1 year old
It's time to retire 'edge' from our IT vocabulary
The term has become so ambiguous it's verging on irrelevance
Opinion What exactly is the edge? What makes something an edge appliance? These are trickier questions than you might think, and depending on who you ask — and honestly, what they’re trying to sell you — the answers can vary wildly.
Yet, the edge is often talked about as if it’s one place or thing. It’s not. At best, it’s a catchall for compute and networking equipment deployed outside the confines of a datacenter.
Today, the term usually refers to compute resources located in close proximity to a data source. The approach allows information to be processed closer to the end users, which has the benefit of improving round-trip latency and reducing the bandwidth that would otherwise be required to pipe it back to a central location. Because of this, edge compute is often cited as a key enabler for several emerging and established markets like 5G, IoT, robotics, transportation, and manufacturing to name a few.
While it might seem like a fairly straightforward definition, where things get complicated is when you start using the term to describe equipment or services.
What the heck is an edge appliance?
In a datacenter, the operating conditions, like temperature, humidity, and air quality are all carefully controlled in accordance to industry standards. You don’t have to worry nearly as much about whether a piece of equipment, like a server or switch, developed for one datacenter is going work in another. But the same can’t be said about the edge.
Depending on where your data is, the edge could be anything from an industrial warehouse to a remote oil rig or a temperature-controlled closet at the back of a department store.
Over the years, modifiers have cropped up to help describe which edge a product or service is designed to serve. There’s the network edge, telco edge, near edge, far edge, industrial edge, edge datacenters, and the list goes on.
It certainly doesn’t help that marketing departments have a habit of shortening everything to edge anyway.
- Intel is running rings around AMD and Arm at the edge
- Edge compute, AI on track for meteoric growth – or so these predictions say
- MIT boffins cram ML training into microcontroller memory
- Singapore branches out onto internet of trees
Glancing through the major OEM’s webpages, you’ll find a smattering of edge appliances ranging from full-fat servers to industrial PCs designed for harsh operating environments. The trouble being that a system designed for an edge datacenter is going to look and function completely different than something designed to work in an industrial setting.
For example, a sports arena might deploy a rack of servers to transcode large volumes of streaming video onsite to save on bandwidth costs rather than piping it back to a central datacenter first. While, this would generally be considered a near-edge application, it’s unlikely that specialized equipment would be required.
A system bound for a manufacturing plant, on the other hand, may be required to operate at temperatures in excess of 85C or in high humidity with minimal airflow and limited power for weeks or months on end. These are not conditions conducive to your typical server, and so a special designation, like “rugged” or “industrial,” is not only warranted but arguably a better description than calling it an edge appliance.
Customers buy equipment to solve a problem, and when everything is labeled the same way, it only serves to make things more confusing and harder to find the right kit. There’s a reason why we describe servers by their function — storage, compute, GPU, virtualization, etc. — as opposed to where they’re deployed.
The edge is overblown
It’s not just confusing. There’s a case to be made that, in the age of the cloud, edge as a term has become so ambiguous as to be irrelevant.
Edge as a definition made sense when the majority of our compute capabilities were locked away in, at most, a handful of centralization datacenters around the world. Today, that’s no longer the case. Workloads can and are deployed everywhere, whether that’s in an on-prem datacenter, cloud region, colocation facility, as a worker on Cloudflare or Akamai, or on a cluster of Intel NUCs at the back of a store.
We as human beings like to categorize and classify things. However, to point to something and calling it an edge appliance or location is about as enlightening as describing an organism by its taxonomic kingdom. It might be an accurate description, but it’s not nearly specific enough to be of any use.
So, while the need to put compute resources closer to data or users isn’t going away anytime soon, it might be time to retire “edge” from our vocabularies in favor of something a bit more specific. What that’ll be, only time will tell. ®