The Linux Foundation recently published its State of the Edge 2021 report. As part of the report, a few leaders in edge technology made some insightful and humorous comments.
Alex Reznik is chairman of the ETSI multi-access edge computing (MEC) industry specification group. ETSI MEC was one of the first groups to work on edge computing specifications.
Reznik said, “My personal opinion — we all need to show the operators the money. When it comes to edge computing, they don’t want to see a strategy about what hyperscalers are making in the cloud. They don’t want to hear about the wonderful imaginary world where every car is self-driving and every person wears VR glasses. If they’re going to invest in rolling out infrastructure at the telco edge, operators will need some low-hanging fruit that will allow them to recoup their investment quickly. Operators want to see the money in six months.”
The Linux Foundation’s LF Edge model focuses on two main edge tiers that straddle the last mile networks, the “Service Provider Edge” and the “User Edge,” with each being further broken down into subcategories.
But Dean Bubley, founder and director of the analyst firm Disruptive Analysis, said people have all kinds of ideas about the definition of the edge.
“When I start a conversation about the edge, I always calibrate where people are on the scale of things,” said Bubley. “Some think of the edge as a megawatt data center in a Tier 3 city. Other people think the edge is a milliwatt processor on a sensor. I think the edge has maybe nine orders of magnitude in both latency time and power, about all of which people say, ‘That’s the edge.’ Different magnitudes of edge apply in different conversations. If there is one thing you can count on, it’s that the edge is going to be messy. Everything is going to be deeply inelegant based on a pragmatism that’s messed up by acquisitions and awkwardness around the physical world.”
Among the many "awkward" things to consider about the edge is the hardware. The State of the Edge report says that hardware deployed at the edge has historically been purpose-built for specific workloads, frequently for content delivery networks or the internet of things. But as edge computing grows in popularity, general purpose infrastructure is also being deployed to run cloud-like workloads.
Jeffrey Ricker is CEO of Hivecell, a company that provides a combination of hardware and software for the edge. Ricker said, “A lot of edge data may never even reach the cloud. Cargo ships have gone from hundreds of sensors on board to tens of thousands. We will never push all that data to the cloud over a satellite connection. We need cloud-like compute power on the ship that can run machine learning models. But wait, it gets more fun…. Cargo ships are only in port a day or two at most, at the whim of weather and port traffic. How will you schedule a technical crew to install the hardware? The answer is, you can’t. Data center hardware will not work in this edge use case.”
And there are other, very practical things to consider such as where to physically place edge compute gear. Lots of companies are getting in on the edge action, including cell tower companies such as American Tower, which is deploying edge computing sites on the real estate it already owns.Ridge, a company that lets developers deploy cloud-native applications anywhere users are located. Seelig, who was also a co-founder of Akamai, said, “It turns out the best space, power and connectivity in every geography everywhere in the world is owned and operated by a data center company in that place. The best data centers in Tokyo are owned by Japanese companies. The best data centers in Frankfurt are owned by German companies. However, despite the incredible space, power, connectivity, and virtualization and compute offerings they have, very few of those data center companies are capable of offering the level of managed services that are the hallmark of what a modern cloud native application wants. The future is to bring together existing data centers in lots and lots of different markets, provide each with the means to deploy managed services, and create a truly global cloud.”
Beth Cohen, a cloud product technologist at Verizon, has also been an edge leader for several years.
Cohen said, “Do not make the mistake of thinking of edge computing as just a mini cloud. It requires an entirely different way of looking at infrastructure, applications, networks and most importantly operational tools for deploying and supporting edge services. Verizon has had network edge products on the market for close to four years now, yet we are still uncovering delivery and management gotchas. The devil is in the details, so while it might be tempting to cut corners, it is essential to complete extensive testing to verify that your edge applications will perform as expected under production conditions.”