We’ve all grown tired of all the unending bloviating by experts on various topics, banal and serious, on cable TV, mostly.
COVID-19 has forced us all to sit through more and more online sessions with Zoom and other formats to the point that “vapid” has lost all meaning. Nowadays everybody seems to call them “virtual” sessions instead of merely “online sessions” when in fact we really are there online—just on a computer with a video stream-- and it’s not usually with some nimcompoop avatar or a weird greenscreen backdrop of a haunted house, mountaintop, the newest tiki bar or some other scene.
(Sorry for that aside. Covid social impacts, whether we’ve contracted the coronavirus or not, are making it impossible to stay on topic, but some of us never mastered that skill before the pandemic.)
OK class, time to get to it. Today’s topic is: “What is the intelligent edge?” Yes, admittedly a brainy topic, but bear with me here.
Today, everybody in networking or computing seems to be talking about the EDGE. Not the edge as in, we are fast sailing toward the edge of the flat ocean, but something else entirely.
Seriously though, the concept of edge computing seems to have started long ago with more broadly accepted “edges” as in what architects mean by it: edge of the forest, edge of the city, edge of a building footprint.
There is also the edge of the wired network out in rural areas and Native American reservations where many people have astoundingly limited Internet access. Wireless networks also have edges, which can be depicted with this actual scenario while using a slightly frantic edge (!) to my voice: ”I am out in the desert and my car has run out of gas and I have no cell service. I’m afraid to turn on my cell phone because the search for a network will burn my battery before I can reach AAA or my wife since she always has me on voice mail and never notices when I text her.”
Anyway, a computer out there in the hinterland or floating on a godforsaken oil platform, or even a specialized computer called a network switch, used to be called an edge node at one time more than decade ago. Or it was just a “node,” often in a network.
There is also that concept of the edge of the power grid. This might be a forest where power lines aren’t strung, for example. Wireless communications can reach into these areas, but wireless power ain’t there yet. Batteries are being developed to serve in all these situations, some of them huge ones that power big machines and facilities and homes and others that are as small as a tiny coin to keep a temperature sensor or other sensor operating for years.
Always a reliable source, Gartner has posited this description of edge computing (not edge networking) : “Edge computing represents an emerging topology-based computing model that enables and optimizes extreme decentralization, placing nodes as close as possible to the source and sinks of data and content.” (This is a two-year old definition, at least, and you can ignore the reference to “sinks” of data and still get the idea.)
My colleague Karen Field and I tried to have a little fun with this question, “What is the intelligent edge?” for the sake of engineers who want to have robots and cars and tiny devices that sense and monitor millions of other things to undertake creative and commercially valuable tasks, often far away from data centers, cloud servers and power plants, but more importantly—near to the source of data. Engineers will be putting artificial intelligence at the edge to make inferences from inputs coming mostly from sensors, by relying on powerful and power-efficient semiconductors and software.
One colorful example of the intelligent edge is what John Deere is doing with Nvidia chips and years of research into its “See and Spray” system to precisely deliver pesticides to weeds in fields of crops. Engineers had to first take high res images of weeds under all conditions in the field to create a data base for the AI to infer a weed. (Imagine going around the world to take weed photos for that job.) Then they use high res cameras drawn behind tractors to find the weeds to spray with a precise amount of chemical, covering no more than a postage stamp in size.
To get a good insight into “What is the intelligent edge?” we reached out to some genuine experts.
Ken Wada, president of Aurium Technologies, answered the question with a question. (Engineers do that a lot and so do journalists. Wada is both. ) Note that he mentioned that word “node” again in this exchange:
“If you mean edge as in point-of-use, there are many tradeoffs and design challenges for engineers in this space,” he said by email. “In networks, edge means those devices that connect point-of-use to backhaul. These edge devices are colloquially known as nodes. The actual network point of use case is typically known as a CPE, Customer Premises Equipment.”
Thanks to Wada there is now more agreement on part of the question of the day, at least between two of us.
By asking “What is the intelligent edge?” there is also an emphasis on “intelligent.” This focus might generally relate to running computer elements that use AI and try to infer decisions, as in, “this is a weed; this is not a weed,” etc. But some engineers might only want to have a device that offers up “intelligence,” which is something more like this: “I am a temperature sensor and I am telling you that I’m both up and running and that the temperature is 70 degrees” in the normal range. If the sensor stops sending information, that’s intelligence about a bad battery or maybe a malfunction in the sensor or maybe even the entire remote building disappeared or blew up. (Too many wild plot lines are another sign of Covid’s impact.)
Not to quibble here, but “intelligent” is not the same thing as “intelligence.” I think maybe somebody in a marketing office for a big chip maker meant to call the phenomenon “edge intelligence” but the connection was garbled and we are left with “intelligent edge.” However, in truth, both concepts apply to what is happening on that treacherous and remote oceanscape edge, beyond which there be dragons.
Also, the edge doesn’t have to be in a remote far-away place. It could be a number of machines in a manufacturing plant just behind the offices but far from normal, old-fashioned connections to networks and power. Again, that’s where monitoring of data is kept close to the data creators, like a machine making precision parts that is monitored with a high res camera and the AI decides what images to store based on finding anomalies. Or, the edge could literally be that big concrete drop-off between the offices and the plant that the executives never seem to want to jump across. Maybe edge is less a place than a state of mind-- an engineer’s mind, to boot. (Deep, eh?)
So once again, we see that engineering is just like every other specialized occupation in that words often mean different things, depending on the situation or scenario. Is some tiny device at the edge truly ‘intelligent” or just offering “intelligence”? That’s a question for engineers, along with these: Can we afford the power needed here? Do they really make batteries for this type of application? Won’t a pack of squirrels eat through the rugged battery case and kill the data stream? When will we see truly autonomous cars filling the roadways? Why didn’t John Deere catalog all the good plants instead of all the weeds for its AI application? And more important: Is it finally time for lunch? Can I still apply to be an astronaut?
Sometimes these crucial questions come down to word choice and that’s a job for editors, but sometimes engineers win out because they make more money and generally laugh at English majors.
Ok, class, thank you for tuning in to this edge bloviating. Now, back to work.
Ken Wada will appear on a panel with other experts on Wednesday Sept. 30 at noon ET, as part of Embedded Innovation Week. At 11 a.m. that day, Chris Rommel of VDC Research will tackle the essential question, “What the heck is the edge and why should embedded developers care.” Sign up for the free event online.