IoT design needs system focus
For good or ill, the Internet of Things (IoT) is upon us. It's still in its infancy, but the patterns are setting quickly. If we don't rethink the way IoT designs are being done, however, we could be in for a lot of trouble, especially in the industrial sector.
There is no question that the IoT is starting to take off. Wearables, sensors, predictive maintenance, production line analytics, connected cars, asset tracking, smart meters, smart homes, smart cities, and smart buildings are popping up everywhere. But all too often the design behind these devices is not all that smart. It's clever, it's innovative, but IoT designs are also all-to-often piecemeal and rushed to market. What's being created is a system of systems, without the system-level design issues getting addressed.
A prescription for correcting the IoT's trajectory, at least for industrial and other critical systems, is coming out of the Object Management Group (OMG). The OMG is a technology standards consortium that has been helping bring together disparate networked systems for more than 30 years, starting with creation of the CORBA (common object request broker) standard making networked software objects interoperable. It has since created such standards as the data distribution service (DDS) and the unified modeling language (UML). The OMG turned its attention to the Industrial IoT (IIoT), taking on management of the Industrial Internet Consortium (IIC) to help define a common IoT architecture and interoperability standards for industry.
The prescription stems from the development of OMG's profile for mission-critical defense systems, UPDM, through its extension for more general systems, the Unified Architecture Framework (UAF). The UPDM co-chairs, Matthew Hause and Graham Bleakly, spoke with me recently about the problems with current IoT design approaches and their suggestions for improvement. "We're trying to get people away from building the IoT by hacking," said Bleakley. "That's happening because of the way the IoT is being pushed, with the idea you can connect anything to anything and it will be fine. The trouble is, in safety critical, medical, industrial, and other systems, if you do it this way you can get into a lot of trouble."
The idea, according to Hause and Bleakly, is to consider the IoT design in light of the entire system, using model-based systems engineering to define the system as well as the design's requirements and to describe how the bits are to fit together. "This approach uses math notation to describe the system. Such models yield more precise specifications and standards for the system design," said Bleakly. "It removes the ambiguity that can occur with just text." With the system models defined, developers can then use simulation to test and refine the models and requirements.
One advantage to this approach, Hause pointed out, is that it provides levels of abstraction that separate the what from the how. The approach uses layers, such as definition of the enterprise objective for the IoT design, desired outcomes, and performance metrics for ensuring the objective has been reached. This model layer is separate from the implementation details, allowing the requirements to be tested independent of the solution. Other layers might include how to handle the data the system generates and how the system is to interact with people such as users and product support. "You start by defining what you need to do," said Hause, "then decide what things go where."
Another advantage of the method is that it can handle the full scale of an IIoT. "Getting the IoT to work at an enterprise level is not a small task," Bleakley said. "These systems are becoming too complex; they don't fit into a designer's head any longer."
Not everything in the IoT can support such a formal and thorough design process, though, Bleakley noted. "This is for systems that absolutely have to work," such as automotive, power generation and distribution, and medical systems. For the industrial IoT, then, taking the model-based systems of systems approach to design makes a great deal of sense.
Oriiginally published on Embedded's sister site, EE Times: "The problem with IIoT design."