IoT – Data Model

by Uwe Meding

IoT platforms on the market today are made to handle the hundreds and perhaps thousands of smart products. Often an afterthought to the main development activities, the data modeling phase is actually so critical that it can cause a technically perfect products to fail and leave businesses ill-equipped to deal with a full commercial roll-out.

Each smart product is unique, with varying characteristics that generate different kinds of data. Even sibling products in the same product line will have as many differences as similarities. It is not a one-size-fits-all situation now, and as the smart products market evolves, greater diversity is highly likely. While vertical IoT solutions for a specific sector may appear to offer a convenient shortcut, they have severe limitations, like being limited to certain types of connectivity, products or use cases. But what if your product doesn’t quite fit? If its attributes are unique and not shared by other product manufacturers in your industry, you cannot capture data about it, or from it, using a rigidly defined database structure.

Customization

Data, it has been said, is the oil of the modern economy because it plays an equivalent, indispensable role across all industries and market sectors. The value to your business of smart products IoT data is equally essential. As a consequence, your IoT platform should employ a data model that enables you to semantically describe your smart product and its multitude of possible interactions in a totally customized way. You’ll need to collect a certain type of data today, and in new kinds of data the future as your product line and your customer needs evolve.

If you want your products to work with the full variety of chip vendors, tagging technologies and even the emerging world of smart labels and packaging (think of a washing machine that gets smarter by recognizing the clothing and laundry detergent that go in it), you need to know that the underlying data model will support any of your product technology choices.

Acquisition

Few IoT devices have some form of user interface, in general IoT devices are focused on offering one or more sensors, one or more actuators, or a combination of both. The common requirements of any of these systems are that we can collect data from very large numbers of devices, store it, analyze it, and then act upon it.

Your platform must be designed to anticipate a very large numbers of devices. The devices are typically expected to create constant streams of data, which in turn may create a significant amount of data. Thus we have a need for a highly scalable storage systems, which can handle diverse data and high volumes.

Some actions as a result of the data stream analysis may need happen in near real-time, so we also have a strong requirement for real-time analytics.

Web-Accessible API’s

Look for a semantic data model that gives you extensibility. This means providing the ability to add additional, custom fields and models to extend your product’s data profile whenever you need to. Ensure that all these data fields are accessible via Web APIs, to enable you to trigger rules, live analytics or reports, and interact with other apps or systems.