“Our edge devices are flaky,” said the Internet of Things admin engineer.
“What’s happening?” the angry, spiky-haired systems control manager spurted back immediately. “I thought you said programming infrastructure and application controls over new-age, cyber-physical system deployments was, well, ‘bread and butter stuff’ — that was the term you used, right?”
SEE: Don’t curb your enthusiasm: Trends and challenges in edge computing (TechRepublic)
“Hang on,” said Cletus, our sysadmin hero. “Controlling edge-centric microcontroller units can be tough, and things can go out of kilter, but there might be an answer in some new technology coming out of Bordeaux, France.”
“Sounds like a complex refactoring and reengineering job,” said the controller. “How many days do you think it’ll take to bring things back in line?”
“Well, according to our French pals, it’s just one line of code,” said Cletus.
That wasn’t real, was it?
Okay, yes, that was a fabricated conversation designed to illustrate and entertain, but it could also happen — that new Bordeaux-based company is real. Luos is the maintainer of an open source project for edge, IoT and embedded distributed systems.
The project’s aim is to create edge device manageability by making hardware as modular as software. It does this by encapsulating embedded device features as software services and then making them accessible via APIs.
The firm has announced an integration with Freedom Robotics, a control and monitoring software infrastructure specialist focused on modern robotics, that enables edge devices to be monitored and controlled remotely over the Internet using just one additional line of code.
“The integration brings a new level of simplicity and power to any cyber-physical system with the ability to see and control microcontroller units using the Freedom Robotics dashboard,” said Nicolas Rabault, co-founder and CEO of Luos. “All it takes is adding one line of code to Luos, making the device visible and able to be monitored and controlled from anywhere using the Internet — down to the level of a specific MCU.”
Luos has partnered with Freedom Robotics to create what it calls a “bridge” that can display embedded data in real-time without needing to be physically connected to the device.
“Developers can use any feature of the Luos software and use that on the Freedom Robotics platform to take advantage of adding all of its functionality,” explained Rabault. “For example, this could mean monitoring and controlling various features of a robot — arms, sensors, lights — that are built with Luos software. This is particularly helpful when, for example, you’re wanting to see the status of a specific sensor in real time, remotely, from anywhere.”
The capability also makes it possible to aggregate data to help with troubleshooting and root-cause analysis. This in turn helps ensure robot edge device uptime.
Luos: Built on microservices
The Luos team is developing its open source solution using microservices concepts for embedded platforms. This essentially modular approach makes it easier to share and reuse code in project implementations that include embedded and edge systems.
The Luos team insists its project delivers a microservices architecture that provides flexibility for unlimited project possibilities, all running as either a single service on a single node or several services on several nodes.
For example, this could be deployed in an intelligent agriculture system that can automatically irrigate a plot of land, a facial recognition robot that can identify the emotions in a face, or a smart parking system that can identify and monitor available spaces.
“Microservices models have long made sense for edge computing, not the least because componentization makes updates and configuration management more straightforward,” said Jon Collins, vice president of research at GigaOm, an engineering-led technology research and analytics house.
This is because, for example, engineering teams can update a single microservice rather than the whole platform when configurations or other parameters need changing.
Collins warns that this does require the team to be on top of how it manages and deploys individual microservices alongside configurations, but coordinated approaches such as GitOps can help here.
“Microservices approaches enable decisions about what should run on the edge device versus what should run in the ‘core’ of the IT stack,” Collins said. “Pre-processing and basic data analysis may be more efficient to run on-device, particularly if networking bandwidth is at a premium. Getting this right requires an architectural view across core and edge, plus the ability to orchestrate where microservices run — orchestration tools such as Kubernetes are not as flexible as they could be, though this is improving.”
The evolution of microservices
Is there a real change happening here? Are we witnessing a step-change in the way we harness microservices to control edge devices, and if so, does this move fall in line with the wider progression path for infrastructure as code, as noted above?
Sarah Polan, field CTO for HashiCorp’s EMEA region, says yes. She points to what she calls the “parabolic increase” in the number of connected devices requiring edge computing capabilities in recent years.
“Given these trends, it’s reasonable to think that with 5G we will continue to see the evolution of microservices in order to support these devices with the required latency,” said Polan. “As this supporting infrastructure becomes more complex, so does the task of managing these dynamic environments. This is precisely why IaC, implemented in a repeatable, industrialized manner, is critical to scalability and operability.”
Not everyone agrees with the sentiments expressed thus far. Some technology advocates see a more prescribed approach to automation as the killer app factor in this space.
James Sinton is a technology platform engineer at a retail technology company in the U.K. called IMS Evolve. IMS has been working with food retailers around the world for the past 20 years to connect devices such as fridges, lighting and air conditioners in stores with IoT to improve efficiency.
“Whilst it could be true to say that no-click and no-code edge systems will usurp one-click and one-line alternatives in the future, I don’t think the technology that currently exists to do that is usable at scale,” Sinton said. “In an ideal world, organizations will have set themselves up in such a way as to automate as much as possible so that assets self-update and generally look after themselves.”
Sinton continued by arguing that such an ideal world is unlikely to come to fruition.
“In reality, though, a lot of enterprises haven’t done the hard work to set it up like that, and I’d probably attribute that to the fact that many enterprises are spending the money needed to upgrade on the human ‘button-clickers’ instead,” Sinton said. “Software solutions that take that button clicking out of play will greatly improve the cost-effectiveness of scaling solutions.”
Edge computing: Modularity matters
The trends coming to the fore here should not be hard to spot. As the cloud computing industry at large moves towards software-defined IaC and the ability to create an abstracted data-defined management plane for our global IT stack, we are emulating that modular manageability factor in terms of the way we work with edge devices, with microservices acting as enabling champions.
If our spiky-haired boss didn’t believe Cletus could pull this kind of control off in just one line of code at first, then they should wake up and smell the coffee: One-click and one-line controls are becoming increasingly usurped by no-click, no-code autonomic systems where the machine does all the work in the first place.