The Internet of Things (IoT) will radically change our world in the coming years.
Estimates suggest that in the next 4-5 years, 30 billion to 50 billion “things” will be deployed. These devices will range from relatively simple sensors (e.g., temperature, pressure, rain) to complex data collectors (e.g., health monitors, shipping/transportation monitors, chemical process monitors, environmental stations).
Many will be cost sensitive “throw away” devices deployed in huge numbers to consumers. But many will be high value devices, particularly in Enterprise of Things (EoT) applications, and in mission critical situations (e.g., health care, autonomous driving, process control, security, smart cities) where many companies are still formulating their strategies.
I don’t expect most IoT devices to be simple data gathers that send collected information to the cloud for analysis and action. That’s not to say that there won’t be simple devices that generate small amounts of occasional data (e.g., water/power meters).
But I expect the majority of IoT devices to include significant processing capability to enhance their functionality and to maximize the true potential of IoT. Indeed, assuming all deployed things will be essentially dumb sensors sending their data up into the cloud to be analyzed and acted on makes little sense. The vast amounts of data transmitted, and the numerous low level processing functions necessary would quickly overwhelm networks and cloud-based systems. What’s needed is localized capability often referred to as processing “at the edge.”
What do we mean by edge processing and why is it even necessary?
Edge processing is the ability to have an intelligent processing capability located as close to the sensor or device as possible. It provides IoT device access and control, security, and localized intelligence that reduces the amount of data transmitted by isolating and sending only the pertinent data needing to be analyzed into the cloud.
Indeed, this can potentially reduce the amount of data transmitted and the need for processing capacity in the cloud by several orders of magnitude. There are a number of commercial “IoT edge server” boxes, essentially modified PC servers, available from multiple vendors that act as localized intelligent controllers and processors. But is this the best way to deploy edge processing?
In many cases the answer is no. Embedded processing “at the source” is often more effective.
Let’s assume you have a camera set up to detect an intrusion of some sort. Rather than sending the entire raw video stream to a remote system for processing, it makes far more sense to do as much processing at the local device as possible. In our case, this would include some form of image processing to discover if there is a moving object in the field of view, and what that object might be.
You might want to send an alert if a human is moving, but not perhaps a dog or cat. A video signal processor built into the camera enables only sending true alerts to the cloud and not random data taking up needless bandwidth and cloud processing resources. It changes the device’s role from video streaming to intrusion detection -- a much higher value capability.
There are other examples of why these things need close-in processing. Intelligent thermostats that control the ecosystem in your home should be programmable and operate autonomously, unless there is some anomaly that requires intervention. Voice controlled light switches should be able to process on/off commands locally at the switch and not need to send audio into the cloud. Drones should be able to act autonomously if you lose control, by including a geolocation system (GPS) and other sensors to make autonomous operation possible. Medical or sports monitors should be able to process information at the source of the sensors and provide alerts and react to anomalous readings at the local level. And virtually all systems require onboard security and privacy mechanisms to avoid data breaches or tampering. There are many such examples of the benefits of making IoT devices smarter.
The examples above imply that each "thing" has the equivalent of an edge processor embedded into the device. With the cost of intelligent controller chips dropping into the few-dollar range due to massive scale in the mobile phone market, this is indeed possible. Mobile chip companies (for example Qualcomm, ARM) have put many of the subsystems we commonly see in smartphones (e.g., CPU, GPU, DSP, image processing, audio processors, wireless) into devices targeted specifically at IoT applications.
Granted the variety of devices to be deployed will require various levels of processing, from limited to high end. But with such a variety of edge processing capabilities, IoT devices can include many higher level functions that move them from dumb sensors to intelligent, programmable, and task-specific analyzers and action enablers, and often without dramatically increasing the device cost.
What does this mean for the future of IoT? I expect that most successful IoT devices will be intelligent and task specific devices that include some level of edge processing within the device itself. There may still be need for an additional stage (or multiple stages) of edge processing before ultimately going into the cloud, as many complex operations (e.g., assembly line controls, plant engineering, advanced robotics) will require more processing power than is available in the multi-purpose chips within the IoT device, or needed for data aggregation of multiple IoT devices.
But with the cost of an intelligent controller adding only a few dollars to the cost of an IoT device, and given the benefits in reducing data transmission and cloud–based resource requirements, its highly likely that most IoT devices will have intelligence built in through the use of edge computing, rather than simply utilizing a dumb controller with no inherent signal processing capability.
If you’re about to acquire (or design) an IoT device, you should make edge processing an important decision criteria if you want your device to maximize its potential.
This article is published as part of the IDG Contributor Network. Want to Join?