The classic view of the Internet of Things (IoT) future is that all sorts of environmental, traffic and other sensors would be connected to the Internet to create a vast new resource of information. Package movement could be monitored in and out of trucks, stores and even homes. Cars would steer their way through IoT-enabled sensors that followed our every move under the watchful eye of IoT traffic cameras selecting the best sensors to track, and the trusted ones to believe.
However, you can probably see by now that this sort of thing would present such a monumental risk to society that it would never stand a chance, even if we got past the basic issues of privacy.
Already today, we have a sea of sensor data that we’re floating in. If you want to know if traffic is bad on your route to work, there are sensors that have that data. But if you don’t know anything about which sensor does what, how can you know which are along your route, or the difference between a “real” one, and one that’s just trying to mess up your commute?
The point is that users, workers, most everyone sees the IoT not as sensors, but as analytics. You see, IoT isn’t even about the Internet, it’s about big data.
IoT & Analytics
In what I’ll call an Analytics IoT (A-IoT) future, companies and organizations would provide sensor knowledge, meaning that they'd collect sensor data and index it so that it could be useful. These providers would responsible for two things: authenticating their sensors so there’s no significant risk of spoofing data to distort results, and authenticating users according to rights and legal constraints. This process would only work if we assumed that the sensors were not directly addressable on the Internet, on closed subnetworks or private IP networks -- something like VPNs.
Users of A-IoT would subscribe to one more of these providers’ repositories and have the right to send queries that would extract sensor information in some form. It’s not difficult to see that the providers would likely differentiate themselves more by the ways they could use context to refine their results, than in their sensors or locations.
For example, who would care if fifty cars passed Sensor X in ten minutes if you didn't know where that sensor is relative to where you’re going to be? Instead, an A-IoT provider could offer a route-to-sensor mapping, where a GPS route was converted into a sensor data set. This would give the requestor something relevant to look at.
IoT & Networking
So is all the "IoT-changes-the-network stuff total nonsense?" In many ways it is, but there are profound issues being covered up by all the “everything-generating-traffic-and-Internet-connections” hype.
One issue is that IoT is a major source of incremental demand for ad hoc analytics that only cloud computing can support. Thus, we can expect a tremendous increase in the use of the cloud, and best of all, a lot of work on application and network architectures to support big data applications that talk to users everywhere, and to each other, on an almost instantaneous basis. That’s the big change, of course.
While there is a lot in A-IoT that will work on historical data -- today we have real-time process control based on short telemetry paths from sensors to machines -- the big issue in A-IoT will be managing the total latency budget of applications. That means not only controlling the path delays between sensors, repositories and users, but also the process latency of both updating and accessing the data. In fact, A-IoT could be a big driver of direct optical coupling of sensor owners, A-IoT repositories and data consumers.
IoT & Practicality
If we can index sensor data by what’s collected, where and when, we’ve come a long way toward making random sensor outputs into something an application can analyze on behalf of a user. What is necessary, and what should have always been a part of IoT planning, is a kind of “functional directory” where information resources can be tied back to owners of sensors and then onward to the sensors themselves. This directory would also insulate applications from changes in sensor type or location, which is essential if we don’t want every change to the sensor universe to make the applications already written obsolete.
Last year, we heard a lot about IoT, and how partnership, security and privacy are critical for IoT success. But, what about utility and practicality? The best guide to effective technology planning is the technology successes we’ve already experienced.
Today, we already have sensor networks. We have big data. And we obviously have the cloud. If we had to start from scratch with someone who had never heard of IoT and present them with a question of how to integrate location/environment information into applications, I think they’d exploit sensor network, big data and cloud just as I’ve suggested A-IoT would.
I also think that until we take this analytics-centric view of IoT, we’re spinning our wheels and putting a very exciting and useful future at risk.
— Tom Nolle, President/Founder/Principal Analyst, CIMI Corp.