All you need to know about the Internet of All Things
Bell Labs released a book last year on the new connectivity era we are entering. Here is my summary of the book including insights from conversations with Marcus Weldon, President of Bell Labs
The goal of the Internet of All Things is to build a digital infrastructure that predicts and automates the mundane, enabling people to concentrate on more productive and creative things. We do this with an IoT infrastructure that puts sensors on everything (animate and inanimate), connects the sensors, thinks about what the things are saying, analyzes the data generated by the things to create knowledge, and finally makes predictions or takes action based on that knowledge. In this way, new value is created in the form of time — time that enables people to be more creative, more productive and more human.
Inspired by conversations with Marcus Weldon and his speech at the NGP World of Connections conference in December, I felt compelled to share these important insights further. There is no other factor as critical to making IoT come to life as the Network of the future. Here is an attempt to explain why.
The “Information Revolution?”
We need to recognize that this is not a normal era. Rather, we are on the precipice of a 100-year shift — the next “technological revolution.” Arguably, there have only been 5 or 6 technological revolutions in the modern era. Each has been characterized by two things: the interconnection of new systems and technologies followed by a profound transformation of economies and society. Looking back at the financial revolution of the 1600s as an example, it was purely enabled by the “technology” of stocks and bonds, interconnected into a new banking and stock market infrastructure that transformed economies. The first industrial revolution of the late 1700s was enabled by steam and iron, connected into rail and shipping networks, transforming almost every aspect of how and where people worked and lived. Technology + Interconnection = Profound Transformation.
Finding value in IoT
Many economists argue that the information revolution has not produced a material increase in productivity, but rather have just broken things apart and moved them around. Measured as the yearly growth in output per person or output per hour, productivity has shown much lower growth rates in the last 45 years than in the 50 years prior. By these measures, we seem to have traded monolithic systems for more distributed systems, without necessarily creating new value.
Today we’re at the precipice of a technological revolution that will enable us to realize significant new value, but we will need to measure productivity differently. If we measure productivity as the sum of human productivity, plus software and hardware robotic/automation productivity, where that automation assists human efforts, productivity will most certainly increase significantly in the age we’re building.
Focusing on the network
“In order to get to an IoT future that can actually deliver revolutionary societal and economic change, we need to completely re-architect the network, moving from a centralized networking architecture to one that is massively distributed.“
To do this, we need to do two things. One, we need to move from the current human-centric level of network latency of 100 ms, to delivering 1 ms of latency. Two, we need to get from a maximum data transfer rate of 10 Mbps per session to handling 1 Gbps. If we can do these two things, we can predict the future with close to 100% accuracy.
How can Marcus make this assertion? Because these two things have only two dependencies: the speed of light, which defines how fast light (communication) propagates over distance, and Shannon’s Law, which defines how much error-free data can be transferred over a channel, given that channel’s bandwidth and signal-to-noise ratio.
100 times lower latency
Our current core cloud network is designed for a latency requirement of about 100 ms, which is more than sufficient for most of the human-centered tasks we perform on the networks. Most human responses — seeing, hearing, blinking, etc. — are quite slow. A sprint runner’s start is among the quickest, taking about 120 ms. This is why so many Internet-based functions — web browsing, chatting, video streaming, even VoIP and high-resolution cloud gaming — work just fine with latency of 100 ms or more.
By contrast, there is one human response that happens almost instantaneously — the vestibulo-ocular reflex that stabilizes our vision, enabling us to move our heads in any direction while keeping our eyes on a particular point. The human response time for this function is 7 ms. Because of this, all networked VR and AR functions will need to respect this requirement, or risk having people getting severe motion sickness. Other network services that require extremely low latency (1 to 10 ms) to be feasible include 360-degree video, cloud-assisted driving, autonomous vehicles and mission-critical automated control systems manufacturing, robotics and power girds.
This is driving us to a network latency requirement of 1 ms. Since latency is dependent on the speed of light, which takes 1 ms to travel 100 km, the answer for both VR/AR applications and the virtualized network is the same — we need to build a cloud infrastructure that hosts the network itself as well as our VR/AR and automated control applications that is no more than 100 km away from every end user, everywhere on the planet, in other words move away from monolithic centralised solutions to flexible local virtualised ones.
100 times higher bandwidth
The other requirement for networked VR/AR and autonomous systems is that they need massive amounts of bandwidth. In 10 years, we predict that we’ll need 60 times current peak bandwidth, dominated by VR/AR applications, — at a minimum, several hundred Mbps
Fundamentally, new applications that will drive IoT into the future by automating our whole world need either very low latency, or very high bandwidth, or both.
The current network technologies can’t meet these needs, or at least not both of them. Low-power wireless access technologies like LoRA and NB-IoT are only good for long-range, low bandwidth access. WiFi supports medium bandwidth at short range. LTE supports medium bandwidth at medium range. But we need to do long, medium and short-range access, and we need to support low, medium and high-bandwidth data transfer. 5G is the only thing designed to cover the entire range of needs, becoming a key trigger for the Internet of Things. The other network technologies will still be important for specific niche markets, but
It’s 5G that will bring us the Internet of All Things.
Our current network capabilities were built to support hundreds of concurrent users and 10s of Mbps for data transfer, but this is only a small portion of what we need to enable IoT. 5G is being designed for our future needs, where IoT adds significant value to our lives. It will include ultra-broadband, supporting the hundreds of concurrent users and gigabytes per second of data transfer required for VR/AR applications. It will also include ultra-narrowband, to support communication with tens of thousands of concurrent “things” that have variable bandwidth requirements, depending on the application. And it will include ultra-low-latency, involving fewer users and lower bandwidth requirements, to support mission-critical system control applications. In essence, 5G is being architected to be the “grand unified network architecture” for everything.
A new architecture for the Internet of All Things
What we have done so far is build a network architecture optimized for humans using devices that cost $1000 with 1-day battery life connecting to a centralized computing infrastructure, whether it’s for messaging, communications protocols or web services — 100s of milliseconds latency, 10s of Mbps data transfer and about 10 billion devices connected, none of which are doing anything particularly revolutionary. For our new IoT architecture, we need to shift to 100 km for the cloud control plane, and 100 meters for the 5G air interface on the currently unused millimeter wave spectrum. This gets us to 1 millisecond latency, 1 Gbps data transfer on that new wave spectrum, 10-year battery life, $1 per device, and more than 1000 billion devices.
We need augmented cognition systems that enable machines to “think” using data analytics and machine learning. And we need the digital value platforms that provide the complex industrial systems that will manage and automate our world for us. We also need a trust framework that provides security at the point of ingress, protecting everything we’re doing on all of the devices using the network.
The goal of the Internet of All Things is to build a digital infrastructure that predicts and automates the mundane, enabling people to concentrate on more productive and creative things. We do this with an IoT infrastructure that puts sensors on everything (animate and inanimate), connects the sensors, thinks about what the things are saying, analyzes the data the generated by the things to create knowledge, and finally makes predictions or takes action based on that knowledge. In this way, new value is created in the form of time — time that enables people to be more creative, more productive and more human.