If you think analytics has been important for telecommunications companies providing a variety of services to hundreds of millions of people, wait until analytics has to be applied to 10 times to 100 times that number of sensors, each representing average revenue per unit perhaps an order of magnitude less than a “human” account.
More than volume is crucial here. So much so that the term “big data” tends not to be so prevalent in the context of Internet of Things, even if IoT virtually requires such analysis.
In fact, there will be so much raw data that many argue that analytics must be applied “at the edge of the network” to discard much raw data before it even is processed centrally.
The thing about many vehicle-related Internet of Things apps is that they require real-time analysis and action, meaning latency has to be quite low.
So a key challenge with IoT is data management.
Determining types of data are important, what should be transmitted immediately, what should be stored and for how long, and what information should be discarded, are key imperatives, according to Mobeen Khan, AT&T executive director.
Otherwise, you could end up with an almost infinite pile of data to analyze, when only a relatively small portion is of real importance, he notes. “Some data just needs to be read and thrown away.”
And much data will have to be processed locally to have greatest value. Systems related to autonomous vehicle and vehicle safety provide examples. Crash prevention is the obvious example.
That sort of data likely must be processed locally, at the edge, often by the vehicle sensors directly. Other functions might be processed locally, but relatively close to the network edge. Apps related to traffic management might be examples. Finally, some data useful for longer-term traffic management can be processed in the cloud, but without urgency.
Over the next five years, AGT estimates, some 15 to 40 billion additional connected devices are expected to be deployed, perhaps a 200-percent increase from today’s installed base.
By 2019, IoT devices will represent an installed base more than double the total of all other connected devices, Business Insider suggests.
Volume is not the only important angle. IoT data’s value will come from decisions and actions based on analysis. Some might characterize IoT analytics as the new customer relationship management.
Gadi Lenz, AGT chief data scientist, warns that equating Internet of Things traffic to RFID or smart meter projects vastly underestimates what IoT is becoming, which can include, for instance, live streaming from hundreds of cameras around a city for public safety purposes.
“The telecoms guys are used to talking about the old IoT, in which you talk about smart meters where they sample once every 15 minutes at best, and send out a little burst of data. “Many others…have different requirements.”
And even if many new IoT-specific data networks are proposed, mostly for low-bandwidth, low-power, very bursty apps, not all IoT applications will be of that sort, Lenz seems to suggest.
System-on-chip advancements mean information can be sent directly from the device that has already been processed and analyzed the raw data, right at the edge of the network. Cisco calls that an example of fog analytics or fog computing.
The reasons for such developments are easy enough to understand, once you look at the apps.
Internet of Things apps for automotive use will have to compute and analyze locally, in the vehicle, in addition to using some amount of more-distant computing and reporting.
So not will “more analytics” be required, the places and ways analytics are applied also will change.