From smart thermostats to fitness trackers, the Internet of Things (IoT) is now commonplace in our daily lives. It is not hard to see why adoption has been so widespread. These connected devices collect, process, and share data on the physical world around us to help make our lives easier and better.
Similarly, many businesses are adopting IoT to use data to better understand their operations, make smarter decisions, reignite customer engagement, and reimagine how value is created. For instance, one supply chain management company has deployed sensors on its fleet of pallets, crates and containers that track geo-location, ambient temperature and pressure, among other environmental variables. This transforms the value the company provides to customers – from renting pallets to optimizing supply chain costs based on unlocking data on the remaining shelf life of goods in transport.
With the rapid evolution of low-cost sensors, elastic computing, and data sciences, many industry watchers expect enterprise deployments of IoT to explode. Indeed, Gartner predicts that businesses are on pace to install approximately 4.1 billion connected things in 2018, and eventually 7.5 billion in 2020.
Experts expect all of these developments to generate about 44 trillion gigabytes of additional IoT data worldwide within that time frame. Which drives us to the central question: What is the best technology architecture to adopt to address this explosive data trend? There are three broad options around this: local, cloud, or hybrid architectures. The answer, as always, depends on the use case.
Local IoT architecture
A local architecture employs edge computing, where data is processed at the edge of the network, nearest to the source. According to IDC, by 2019, 45 percent of IoT data will be stored, processed, and acted on close to or at the edge. This model provides a smaller performance footprint, which can help companies make more real-time responses to data. For instance, on an oil rig, sensors can detect if a faulty valve poses a fire hazard. In such a case, enterprises cannot afford delays. If the data needs to be sent up to a satellite, over to a data center and back before a notification goes out to shut off the valve, it can be too late. But with a faster performance footprint, the data does not have to travel far from its source. This reduces time delays and allows for time-critical decisions.
Further, local architectures are not reliant upon internet connections like cloud environments. And local architectures are also favored by companies with stringent data security concerns. There are many use cases where local architecture utilizing edge computing makes a lot of sense.
Cloud IoT architecture
A cloud IoT architecture can be beneficial for organizations managing a large volume of connected devices where value is driven through the combination of internal and external data. For instance, supply chain applications benefit from understanding the specific view of a piece relative to the aggregated view of the whole. And only one set of data outside of the full view loses its significance. For instance, it would be impossible to try to orchestrate the supply chain for each component of an asset build by using local architecture alone.
Additionally, cloud architectures offer greater interoperability in being able to integrate and interact with other IoT devices and cloud systems. This model provides far more architectural flexibility and leverage of external data sources. Finally, cloud applications are seeing the highest innovation in the ecosystem partly because software developers target large marketplaces first. IoT deployments that leverage cloud architectures can be more efficient because of the innovative, and competitive, offerings already available. In essence, a cloud architecture can allow organizations to future-proof their IoT investments.
Often, the best approach is one that efficiently combines processing of large core data sets at the edge and then processing a reduced set of aggregated derivative data at the core. As an example, smart cities that deploy parking sensors can process all the sensor data closer to the garages and bring only summary data on the number of spots open in various garages to make intelligent routing recommendations to drivers approaching downtown. After all, it can be expensive to move all that data around every few seconds, and motorists approaching destinations do not necessarily need to know which exact spots within a garage are open. In such instances, a hybrid architecture is ideal.
Another example from the asset optimization world is applications for wind turbines that use sensors to collect and analyze data on each turbine locally and optimize their collective performance in the aggregate. Here many individual data points together allow a deep view on the health of a turbine’s component. Each component’s health aggregated with those of other related components provides a view of a single turbine. Finally, aggregating the summary from all the turbines provides insightful and actionable information on the wind farm. In a situation like this, how much data should get processed at the network’s edge versus what is brought to the center and processed at the core is an important architectural consideration. The combination of real-time response of a local architecture as well as the cloud’s system-wide access and scalability brings the best of both worlds to play.
Consider business needs
In the end, design considerations can drive informed choices on the data and processing architecture for IoT systems. To determine what IoT architecture is the best fit, look across organization’s current and planned devices, business objectives and context, associated processes, and range of planned outcomes. Evaluate these business needs with technology considerations of scalability, performance, bandwidth economics, and rate of technology innovation.
The opportunity to capitalize on the growth in IoT is massive. As the pervasiveness of IoT-enabled devices continues to expand, both in our professional and personal lives, it is essential that companies not only think about the business models and deployment plans, but, design, fundamentally, what system architecture is necessary to materialize the promise of IoT in their businesses.