The opportunity of the Internet of Things (IoT) has everything to do with data. By distributing and diversifying data, IoT architectures will dramatically impact analytics, stakeholders and infrastructure.
Analytics are highly distributed in an IoT architecture, meaning that distributed management and data use must become the norm for information managers.
“The IoT means massive distribution of data and the processing of data,” said Nick Heudecker, research director at Gartner. “Data will be produced, collected and stored in multiple locations depending on the nature and goals of the IoT architecture and use case.”
Data produced by devices may be stored in the device itself, in intermediate locations or in a centralized repository resident on-premises or in the cloud. The processing of that data may happen in any and all of those same locations. Many devices will be powerful enough to perform sophisticated computation on the data they generate, and/or house and process data locally for autonomous behavior.
Some IoT scenarios will rely on highly centralized collection and processing of data in traditional on-premises environments or cloud-based repositories and compute platforms. Frequently, these scenarios will overlap and may create interesting side effects.
One such example can be found in farming, where agricultural irrigation sensors coordinate with local equipment to optimize water use and crop yields. In this scenario, cloud-based information shared with seed providers improves crops the next season, while farming equipment manufacturers discover unapproved usages that void equipment warranties.
Similarly, in the automotive industry, connected cars engage with drivers through location-based and context-aware services. These same sensors report to centralized stores and are used by law enforcement for asset forfeiture, and insurance companies for near-time premium adjustment.
These examples highlight the distributed nature of IoT use cases. “The characteristics of IoT architectures mean that information management practitioners must swiftly become adept at managing many pieces of information spread over a wider and more diverse landscape of platforms than ever before,” said Mr. Heudecker.
Organizing and managing highly distributed data is by itself a significant challenge; but ensuring the distribution and consistency of business rules are applied to the data, and monitoring the execution of those business rules, adds additional layers of complexity.
Steps for success:
Embrace hybrid architectures: IoT solutions will generally involve a combination of platforms, with data and process on that data being located “on-device” and in traditional on-premises and cloud-based environments. It is therefore important to avoid forcing analytics and information management solutions into a monolithic or “one size fits all” deployment model.
Plan for disaggregation and resiliency: Workloads must be able to be broken into components that can run anywhere. This means that the implementation of existing business logic for processing data, and current choices of data management infrastructure tools, may not be suitable for IoT requirements — driving organizations to rearchitect or modernize those capabilities.
Focus on monitoring and manageability: IoT architectures will often be the opposite of monolithic, increasing the challenges of monitoring and managing distributed data and its consumption. As a result, information infrastructure capabilities that can be located and managed anywhere should be deployed.