How to Manage the Explosion of IoT Data

The Internet of Things (IoT) has evolved quite rapidly, especially if you consider how long it took for little packet-switching networks like ARPANET to morph into today's nearly ubiquitous internet. Our nascent IoT ecosystem is already populated by billions of sensors and connected devices, hundreds of millions of servers, and thousands of hybrid clouds serving billions of people... all of them perpetually producing and/or consuming data in an explosion of information.

This evolution has given rise to entirely new user interfaces influenced by models geared to augment reality and create frictionless human experiences with devices and applications. We now have an expectation of interactive immediacy with technology that is rapidly becoming second nature and represents a big, generational paradigm shift. What we think of as a computer is changing, as is the way we engage our technology.

Consider your laptop or smartphone and all the many applications you use both personally and professionally, along with the many sites and services and sensors that power those applications. Then think of a similar set of apps for a business unit, then a company, then a conglomerate. Add to those all the connected technologies that seem to have grown commonplace overnight:

• GPS trackers for cars and commercial fleet management
• LiDAR-dependent vehicle safety and traffic management systems
• Energy usage monitoring systems and air quality sensors measuring on a room-by-room or city-by-city basis
• Smart outlets that enable remote appliance or equipment control
• Connected HVAC systems, security systems, and appliances that don't require smart outlets to permit remote user control
• Drones deployed by hobbyists, businesses, news organizations, governments, and armies
• Inexpensive speakers equipped with natural-language processing engines and reimagined as digital personal assistants (Amazon Echo, Google Home)
• Smart routers and gateways increasing geared to take on management of other connected or smart devices and services.

The list could run on through hundreds if not thousands of use cases. All of these devices and sensors and applications and their various modes of connectivity surround us - with more coming online every day - collectively creating astronomical volumes of data. IDC estimates that by 2020, the data we create and copy annually will reach 44 zettabytes or 44 trillion gigabytes.

This crush of computing things facilitates our augmented capabilities and new user experiences because the technology that makes them possible (machine learning and advanced analytics and all that now broadly falls under the "AI" banner) requires massive amounts of data to work properly.

Therein lies the problem. The explosion of "things" makes understanding what needs to connect to what challenging. As noted in Forrester's 2016 Internet of Things Heat Map, "Because of the enormous range of sensors, customer scenarios, and business cases, the technologies for IoT sensor devices, radios, network protocols, software protocols, and data formats are very diverse. There is nothing like the market norms and industry standards of data center, PC, or mobile device technology." Bandwidth is also a concern. Cisco estimates that global machine-to-machine connections alone will grow to 12.2 billion by 2020 and that global M2M IP traffic will grow six-fold over this same period. Managing that enormous flow of data - wrangling, cleaning, culling, transporting, storing, exchanging, integrating - is tremendously complicated and expensive in terms of both time and money.

Until now, IoT implementations have relied almost exclusively on the cloud and large data centers for compute power, bound by bandwidth availability. But that ephemeral model just can't continue to hold up under the data-deluged weight of our new reality. What to do?

The solution for information technology architectures is the same as it is for architecting physical buildings of might and scale: We must distribute the structural load.

In the case of IoT, the new blueprint involves shifting a great deal of computation to the "edge" - out to the local gateways and devices themselves (as opposed to having to transport everything to the cloud). Edge microservices and tiny AI engines can now handle application logic, event processing, and integration duties reliably in real time at the router level (often even at the sensor level), then transmit only essential details to the cloud. This is a better fit for emerging IoT use cases. For example, if an autonomous car has to react in a second, its software can't take three seconds to start up. Aside from reducing latency and enabling real-time responsiveness in devices, edge-based microservices also reduce bandwidth blockage and overall expense because fewer, coarser data packets need to travel to and from the cloud. That expense reduction is impressive: Researcher and Wikibon cofounder David Floyer ran an analysis that showed the costs to be about 1/3 of a cloud-only approach. And security implementation also incurs edge benefits. As Paul Stevens of mission-critical telecom hardware firm Advantech, explained in Senza Fili Consulting's 2017 Power at the Edge Report, virtual functions can now be run either on devices that are at the edge of the network or up in the cloud. "The security gateway function... is evolving, especially with densification," he said, "We're finding that it's better adapted to aggregation points at the edge, which secure data as it hits the network instead of placing gateways in the core." InterDigital's Debashish Purkayastha echoed this logic in the same report, "As personal data traverses the internet, it is vulnerable to being stolen. With edge computing, we can process data at the edge, so we can keep [sensitive] data in a local context. We don't let it go outside the local context, into the internet."

The cloud is still an essential tool, but with this edge-computing shift, a new model emerges where constant cloud connectivity is not essential. Most data will be digested at or near its point of origin. Things that require real-time response (as in an autonomous vehicle approaching a red light) will be handled on-site in the moment (in that individual car). Things that require global distribution on an intermittent basis (as with system updates and upgrades to an entire line of autonomous vehicles) can be optimized, managed, and dispersed via the cloud. The cloud's role then becomes synergistic as opposed to directive: Its economy of scale can be directed to build and refine machine learning models on massive data sets, thereby further augmenting intelligence, which can further be acted on at the edge.

This expansion of compute capability at the source of interface with the physical world represents a natural next-step toward IoT maturity, and delivers a strategy for best utilizing our technology in an increasingly data-driven world.