The past year has seen Internet of Things (IoT) evolve from an IT buzzword to a strategic business imperative, as a steady drumbeat of big business projects and vendor product announcements have legitimized the concept of connected devices. IoT was one technology trend that analysts predicted correctly in 2015. About 10 billion connected devices are currently in use, and various forecasts predict that number will double or quintuple by 2020. This translates into at least a billion dollars in annual revenue for companies active in the IoT industry, with a total economic impact rivaling that of the German economy by 2025. Even if these estimates prove wildly optimistic, companies and IT developers can't ignore business applications that promise new sources of revenue, higher customer satisfaction and greater efficiency by incorporating intelligent, connected devices into products, services and business processes.
Consumer products like wearables, connected appliances and smart home controllers have generated most of the IoT buzz, but the more important profit- and revenue-enhancing applications come from adding sensors, intelligence and connectivity to equipment. The combination of smart sensors, cheap, battery-powered processors and storage -- and ubiquitous wireless networks -- yields a bonanza of new information that can be transformed into business insight.
Indeed, "things" are only half the IoT story because device "intelligence" is a relative term: They only collect and distribute data about local conditions with the ability to process the data. Thus, IoT is equally a big data problem because the whole point of connecting intelligent devices is to gather and share data -- information that once aggregated and analyzed can spot trends, detect problems, flag anomalies and modify actions. Yet, IoT isn't your typical big data system because it involves thousands, if not millions, of data sources scattered across myriad remote networks that, combined, can generate enormous amounts of data.
Cisco estimates that connected devices will create 507.5 zetabyes (1 billion terabytes) of data per year by 2019. Although most of this raw data -- like machine telemetry or device logs -- will never make it to a data center, it still implies gigabytes, if not terabytes, per year, per device flowing into some sort of IoT analysis system. The question is where? What can handle IoT data volumes, from millions of connections, where the data flow can be highly variable and episodic, and process the data into useful information? Hyperscale IoT cloud services are a natural fit.
I agree with IDC's forecast that within five years, "more than 90% of all IoT data will be hosted on service provider platforms because cloud computing reduces the complexity of supporting IoT 'data lending.'" IDC also projects that "the growing importance of analytics in IoT services will ensure that hyperscale data centers are a major component of most IoT service offerings." That is, IoT will fuel cloud growth.
We already have an example from the smartphone world. Mobile app developers needing back-end processing, data aggregation and state management for millions, if not billions (in the case of Facebook), of connected clients, recognize the value of cloud back ends and have fueled the rise of mobile backend as a service (MBaaS) products. IoT is following a similar path, although this time, cloud providers are ahead of the app developers.
Intelligent devices generating reams of data are coming whether enterprises want them or not. Industrial and IT products will increasingly provide much richer telemetry about their state of operation, usage and anomalies. However, without an IoT data collection and analysis strategy, organizations will end up wasting it. I offer the following suggestions:
- Investigate the IoT features of existing data center, manufacturing and facilities equipment and select a few areas in which better understanding of operating conditions might eliminate service calls, prevent or mitigate equipment problems, or provide a deeper understanding of user behavior.
- Organizations developing hardware products should make IoT data collection and communication a part of the design. Look at reference architectures from component manufacturers like Intel, Marvell, MediaTek and others.
- Exploit IoT cloud services for the data aggregation and processing back end. Although AWS and Azure lead the way and have beta services available today, others are sure to follow.
- Build the IoT software architecture on three pillars: data streaming, collection and management, and big data analysis and security, looking at the full spectrum of authentication, credential and monitoring features.
IoT is still a new and dynamic field, meaning that projects must start small, adapt and iterate quickly, and include user and business unit feedback early and often because the goal is improved operations, greater efficiency and new sources of revenue. Look for problems that could be fixed easily with better information but that don't require a major new hardware design (unless, of course, you're in the hardware business and starting a new design cycle). Using IoT cloud services removes a major project roadblock, namely back-end infrastructure deployment and management, and will reduce the time between idea and results.
How to ease development with MBaaS platforms
MBaaS tool connects mobile apps, back-end services
Get up to speed on MBaaS