The edge: Can we teach an old dog new tricks?

The edge has been “real” for years. What’s new is the scale and scope of mission-critical edge deployments.

It’s safe to say the data centre industry expects a lot from “the edge”. The short, sharp, edgy term evinces the bleeding-edge and the frontier of innovation.

On the typical view edge is a (relatively) new IT trend, demanding specific, high-tech innovations to process data at remote enterprise sites away from the data centre, caused by an uptick in demand for ultra-low-latency applications, such as those required to orchestrate autonomous cars.

The argument goes that the need for low latency in a growing number of scenarios means enterprises don’t have time to make repeated trips back and forth to the cloud or central data centre. Edge processing in these applications combines monitoring and analytics. Monitoring demands “IoT” type devices to sense critical environments, such as oil and gas rigs, and analytics requires solutions such as edge-based AI inference and training that can analyse and shoot out data back in real time, say, to proactively warn personnel of danger.

As the edge provides an edgy name for edgy technologies from a marketing standpoint it is without equal. The term’s inherent marketability has also led to what you might call EFS (Edge fatigue syndrome): it’s not that architectures are not being deployed in specific scenarios where real-time response is a necessity. The problem is that the law of low latency is not as generalisable as we are told. Can the average enterprise justify the significant outlay on new edge infrastructure to save a few ms, rather than relying on predictive analytics in the cloud?

Last week Schneider Electric welcomed journalists to its Boston facility for a mini-conference named ‘Life at the Edge’. Over the three days, we were systematically guided through Schneider’s edge strategy — encompassing integrated systems, a cloud-based software stack, and an ecosystem of partners — alongside examples of the edge in action. Schneider is in little doubt that a considerable number of commercial, industrial and telco businesses will soon be dependent on all three ingredients. That’s because Schneider has a slightly different take on what the edge is.

Defining the edge

As mentioned, it has been commonly accepted that edge computing simply describes the processing of data close to where the data is being produced. What tied together the wide-reaching presentations and demonstrations across the three-days in Boston is that Schneider views it a mistake to think of the edge as referring to a necessarily new breed of distributed infrastructure. Au contraire. The edge has been around for years. In server rooms, in operating theatres, in research facilities, if you look at the edge primarily through the lens of mission-criticality.

Kevin Brown CTO and Senior VP of innovation at Schneider Electric’s power division, told attendees that the mainstream definition is unnecessarily narrow. The edge, he said, is simply the “first-point you’re plugging into the network” which encompasses any computing enclosure, space or facility physically closer to the point of origin of data or user base, such as server rooms or wiring closets.

Retail

Schneider Electric is focusing a lot on addressing these needs in retail. Brick and mortar retailers are struggling to compete with Amazon by offering unique services that differentiate them from their eCommerce counterparts. These days, retailers can sometimes have thousands of sites that rely on IT, explained Kevin, even if they are just facilitating network connections to the cloud or wiring closets powering in-store wifi. If these sites go down they ‘will fail on their mission to beat Amazon’, he said.

“We talked to one retailer, where basically a wiring closet, so to speak, is powering all of their wifi. They have power over ethernet so it’s powering their phones or the phones that they have. They’ve taken all the servers out, but they’re consuming more power because they’re constantly adding more wifi points.”

“Now as a result, that’s becoming mission-critical, even though there’s no computing there. The question becomes ‘what are you doing about that first point on the network?’ regardless of what’s in it.”

This demands that enterprises treat these sites more like data centres, ensuring that they are secure, highly available and resilient. This does not mean everybody must bestow edge sites with tier three capabilities, rather “crawl up that path” and get resiliency to the appropriate level.

“No one’s coming in and saying, ‘I’m going to put in two utility feeds, and do concurrent maintenance the same way that you do for a tier three data centre.’ That’s not the conversation. It’s about how I manage it better? How do I service it better? How do I deal with physical security, cybersecurity?” Kevin said.

Security

Schneider explained a variety of ways it is dealing with the problem of physical security at the edge. One micro data centre, virtually unmistakable from an office desk, is equipped with vibration detection to alert managers to unwarranted transportation, and a camera behind the door of the enclosure that snaps the face of anybody who opens it. All these threat alerts are passed through to data centre managers via Schneider’s cloud-based software stack EcoStruxture.

“I’ve walked into people’s environments, and seen a rat’s nest of cables. Somebody can walk into some of these stores, and literally, all they have to do is bribe the janitor,” Kevin said.

The security challenge is still far from being addressed – and boy will it need to be if businesses scale out their edge sites as Schneider expect them to. Many enterprises simply don’t have the human capital to protect edge deployments in the same way they can their core data centres, nor the capacity to respond to attacks there. Taking a picture is all well and good, but not, say, if the assailant is wearing a mask.

Another example proved how turning edge sites into data centres can deliver dramatic improvement to the operational and electrical efficiencies of an enterprises’ entire IT estate. Schneider flew in Simon Binley, data centre manager at the charity Wellcome Trust, to talk us through the data centre transformation project carried out by Efficiency IT, a Schneider partner. Although still in the deployment phase, Wellcome worked with the companies to install Schneider’s EcoStruxure software stack, and physical data centres at its edge.

For Simon Binley, his edge is the 25 genome sequencing machines housed in a sequencing facility outside of his main data centre that process genomic data around the clock.

“That’s our edge because that’s where the data sequenced hits the data centre,” Simon explained.

The use case is yet another example of the diversity of edge deployments. In Simon’s case, he actually does need tier two or even three capability. Each machine performs sequencing using complex chemicals that cost £36,000 a pop. Should the machine ever go down, all goes down the drain – literally. 24 hours of yearly downtime translates to £900,000 in operating costs that would otherwise be spent supporting the charity’s scientific research.

Whether you’re a charity or a commercial organisation, it’s not just about “the core” anymore. It’s about looking at the entire estate and all of the businesses major elements. The edge is real, and has to some extent always been lurking most in enterprises’ estates. The scale and scope of mission-critical edge deployments is what is new. To deal with the rising tide, Schneider has rightly anticipated that businesses need new infrastructure and tools, and must forge new and complex channel relationships.