Part 1 in the 3-part “Edge Computing Series”
When it comes to defining IT terms, I say the simpler the better. Let’s start with the Internet of Things (IoT). The IoT is the network of connected things—like industrial machines or coffeemakers, things with sensors and APIs that enable connectivity and data exchange. Simple enough.
Before defining edge computing, however, we need to understand that the “edge” gets its name from its relation to the core. The core is simply the collection of technologies (housed in a data center or distributed in the cloud) that make up the critical IT and business functionality for any organization. When a business deploys an IoT scenario—say a series of HVAC machines throughout a college campus—the machines are, by definition, deployed at the edge.
Another term bandied about is “edge processing”—which is basically data processing that happens at the edge rather that at the core. This brings us to the question: Why process at the edge?
An answer to a problem
The idea of processing data at the edge is a solution to a practical engineering problem. IoT as a concept has always assumed that connected things would exchange data with the core via the cloud. Problem is, obstacles stand in the way. Some of these involve:
- Bandwidth: For many IoT deployments, the bandwidth required to transmit data from edge devices is cost-prohibitive.
- Connectivity: For moving deployments (such as connected vehicles) or for deployments in remote locales (such as an oil rig in the ocean), connectivity may not be reliable.
- Latency: In situations where real-time data is required—say construction equipment designed to detect and avoid potential collisions—the data latency of the cloud is unacceptable.
- Power consumption: Many sensors in edge devices cannot live up to the power-consumption demands required for transmitting data to the cloud.
- Security: Most sensors—often limited in functionality—cannot provide the kind of security required in a digital economy with an expanding threat landscape.
Edge computing overcomes these obstacles with the use of an IoT gateway. Think of the gateway as a hub of sorts that lives in close proximity to the edge devices within a local area network (LAN). This hub—a full-blown server or something more purpose-built—can help conserve bandwidth by running an analytical algorithm to determine the business value of incoming sensor data and transmitting to the core only what makes sense. The hub also addresses the issue of intermittent connectivity by housing software and functionality that can be used to make decisions on the ground without access to the core. Similarly, latency and power-consumption issues are addressed through the hub, which communicates quickly and efficiently with sensors through fast, low-energy protocols such as Bluetooth or ZigBee.
On the security front, an edge computing hub can provide secure tunnels back to the digital core IT infrastructure. Remember that the October 2016, a denial of service attack—which brought down the Internet in North America and Europe—was executed with a botnet of unsecured IoT devices. As far as use cases for edge computing go, let’s call this one a slam dunk.
The role of microservices
Microservices as loosely coupled, independently deployed nuggets of application functionality. Communicating through APIs and running as unique processes, microservices are ideally suited for IoT scenarios. Why? Because they’re deployed in isolated containers so that if they fail, they don’t take down the entire network or interrupt an entire business process.
The idea is that microservices are created at the core and then delivered to the hub at the edge. The hub then makes them available to each device on the edge as needed. The algorithms that determine the business value of sensor data? These run in microservices. Predictive analytics—let’s say to predict the failure of an HVAC machine on the college campus? Also delivered via microservices. Indeed, microservices are what make the IoT a practical reality. Without them, the IoT would still be only a concept.
There’s a lot more to discuss on the topic of edge computing, but I’ll stop here for now. To dive in further, see this paper on the “4 Ps” of intelligent edge processing: Excelling at the Edge: Achieving Business Outcomes in a Connected World. Also look for my next blog in this series: “The IoT Data Explosion, IPv6, and the Need to Process at the Edge.”