IoT edge devices are finally bringing cloud capabilities directly to the users and devices that need them. Here are the five elements of IoT edge devices.
The Internet of Things promises to reshape much of how we live and work today, whether we’re changing the temperature of a thermostat at home or waiting at an IoT-enabled traffic light. Connected sensors and the data they generate promise a set of experiences that were largely the stuff of science fiction a decade ago.
Between 2016 and 2021, IoT edge network spending is predicted to grow 30% to reach $2.1 billion, according to IDC, a research firm.
Internet of Things- (IoT)-enabled provide rich insights into the health, performance or status of a given device and the device’s surrounding environment. Data insights can automate changes and make environments work more efficiently.
When IoT data is gathered from multiple devices, it can help predict failure and prevent costly maintenance issues, improving worker safety, reducing revenue loss due to product defects, even saving lives. Most IoT systems require cloud computing to unlock the capabilities of the platform, including rapid scale, high-volume data storage, and machine learning to identify critical patterns in the data.
Here’s how an IoT system works: An IoT device or sensor generates data that is sent through a messaging system, evaluated against complex event-processing rules, and then stored for downstream applications, reporting, and further machine learning. However, in many critical business systems, there is another important role: an IoT edge device or gateway. IoT edge devices enable data to be gathered and processed at the edge, where the device and users reside.
In this article, we explore IoT edge platforms, the architecture of these systems, and their future.
IoT edge devices solve a fundamental problem associated with the centralization of cloud architecture. While clouds are powerful for storage and processing, they create delays for IoT devices sending data back and forth. By bringing cloud computing capabilities to local devices, IoT edge computing can process data faster, preventing delays, security breaches and other concerns.
Local processing supports real-time business systems and alerts, and it manages the data sent back to the cloud. Consider an autonomous vehicle traveling down a busy road. If the car needs to stop immediately to prevent an accident, sending data to the cloud will likely take too long. Instead, the thousands of sensors that assess the status of every piece of equipment, need to be able to manage the data locally and respond in fractions of a second. Similarly, a machine failure in a manufacturing plant could result in defective products, delays in processing causing significant financial loss, or even loss of life.
In these scenarios, an IoT edge device meets the needs of the larger IoT platform while also bringing decision-processing local by eliminating the roundtrip time required for cloud processing.
Like IoT platforms, several vendors offer IoT edge technologies and devices. While these offerings vary in services and capabilities, there are five core features that exemplify these systems:
In addition, all edge devices offer numerous types of networking that support options for Ethernet, wireless, and, soon, 5G—the next-generation connectivity for wireless networks. Further, the edge devices may come in hardened and durable cases designed for harsh environments or may appear like a small consumer electronic. Regardless of the physical casing and the specific network options offered, the software and services supplied on the edge device are often the most important factors in deciding what is needed to support the business.
The five components of IoT edge computing
1. Complex event processing. Complex event processing (CEP) software and services have been used for decades in many operational technology deployments. A CEP system takes in data from multiple sensors or signals and then acts on the data as specific patterns arise. In an edge platform, CEP models and pattern matching are developed in the cloud and then pushed to edge devices. One of the common open source technologies for building and running CEP systems—in the cloud and in edge devices—is Apache Storm.
2. Machine learning (ML) and artificial intelligence (AI). The core idea of machine learning is to enable a machine to learn the significance of data. These models may be built using known algorithms, while more advanced machine learning tools focus on training the computer to learn patterns and anomalies so that it can then learn on its own.
As ML models become more sophisticated, they can be turned into AI technologies. A search on Google for images with a beach, family, and a dog will yield only images that match these words. As Google processes an image, machine learning models learn what a beach is, what a dog is, and what a family is. These models are constantly refined with additional data points.
Most IoT edge devices now support machine learning models locally within the device using TensorFlow or other languages. Some IoT edge devices go further and deliver AI technologies directly on the edge. These AI capabilities are also now delivered directly to IoT devices themselves. For example, one of the major video doorbell devices provides AI directly on the device, enabling the service to announce the person at the door.
Running ML and AI models directly on IoT edge devices enables capabilities to be handled directly on a device. Localizing the data reduces the latency that results from sending the data to the cloud, and it enables more immediate insights generated by devices.
3. Applications. Some of the more powerful IoT edge devices have introduced the ability to run applications directly on the edge device. Most of these platforms have been designed to support containerized applications, allowing specialized apps and services to be delivered directly to the device. Some vendors have packaged all the software capabilities into containerized app models, including the CEP, AI, and other tools.
This approach enables an edge device to be dynamically built based on the business needs, location, and other business rules, while providing a quick way to update or change the device.
Applications become an important element of IoT edge devices as data, decision processes, and alerting / monitoring systems are now running directly against the data and completely local. Consider the autonomous vehicle example provided earlier. CEP, machine learning, and AI models are constantly being used to power the vehicle. The apps that form the foundation of how the car operates and how decisions are made could be a small set of containerized apps running directly on a small IoT edge device located somewhere in the vehicle.
4. Offline data. In many IoT deployments, the network connection suffers from constrained availability. In these cases, many IoT edge devices include storage options to hold data temporarily until a connection is restored. Using the local apps and services on the IoT edge device, decision systems can be supported without the connection to the cloud.
5. Data management. The final element of the core IoT edge device is data management: knowing which data to keep and which to discard because it has little business value. In addition, data management may provide data aggregation. Suppose a sensor was installed on a door. The sensor might indicate whether the door is open or closed. In addition, it might indicate whether a door is stuck, blocked, or several other statuses. After enough data has been collected to understand the failure, a door-open status may provide little ongoing business benefit. In this case, an IoT edge device could aggregate the data giving an overall health of the door and reduce the number of “door opened successfully” data points that are pushed back to the cloud. Consequently, this reduces the amount of data being sent, reduces the strain on the network, and offers ways to support the remote and more disconnected sites where bandwidth is constrained.
In addition, in many deployments an IoT device may require support for an older or non-standard protocol. In these cases, an IoT edge device can provide features to help manage the data that is collected, translated and delivered to the cloud.
In the future, a key aspect of IoT edge devices will be flexibility to rapidly change device configuration through remote tools as well as to deliver new capabilities and applications. Further, all packages, updates and controls will include increased security to block attacks while driving more automatic deployment. The goal is to enable a technician at a local site that has no background or understanding of IoT and IoT edge devices to attach a power cord and a network cable, then walk away with the device authenticating automatically and self-provisioning. If the device needs to be moved, it could automatically self-provision to the requirements of the new location.
While nascent, this future of IoT edge devices is emerging. Vendors that evolve in these capacities will also invest in more specialized apps and services native to the device to enhance its edge computing capabilities.
To get started, consider the needs of the IoT application, the type of data collected, if any protocol translation is needed, and the types of services that are required. Further, consider what the specific IoT platform provider supports. These data points will then inform the decision process on the type of IoT edge device that should be considered. Once the software requirements are known, consider the type of network connections needed, the environment, and the physical requirements for the IoT edge device.
Most IoT edge devices will offer similar capabilities and generic services to appeal to the broadest set of business cases. However, most IoT edge vendors will offer specialized devices that place a higher priority on memory, support for more container management tools expanding the applications supported, or including specialized processors to assist with AI workloads.
Consider a business application that processes image data from multiple IoT-enabled camera systems within a building to identify people in the facility. The app sends the imaging data back to a cloud-based platform, and provides real-time alerting should people be identified. These then become the requirements for the IoT edge device. In this case, the device needs to have larger storage capacity to store the images and support the potential loss of the network. In addition, the device needs to include larger amounts of memory to support the processing of images. Finally, the device needs to offer support for CEP to take pattern matching from the data flowing into the IoT edge device and then trigger alerts.
This kind of scenario illustrates the power of IoT-enabled devices at the edge.
Sean Bryson is a high-tech executive and consultant focused on innovative technologies.