The Internet of Things: what, who, how and why

The IoT is poised to change everything. Here are the facts

The Internet of Things (IoT) is all about extending the functionality of the familiar internet far beyond the confines of servers, PCs and smart devices to billions of "things" that are for now totally unaware of their surroundings - household appliances, cars, doors, vending machines, and so on. By fitting them with tiny sensors and actuators, these passive devices can be made "intelligent", able to react to events autonomously rather than having to rely on direct input from their human masters.

However, in most cases, plugging these things straight into the standard IP network infrastructure is an impossibility. Running on batteries, the tiny components are severely constrained in terms of the amount of energy they can use, and the memory and processing power they can bring to bear. Moreover, the radio frequencies and bandwidth available to them is restricted by law.

Huge leaps are being made in many of these areas, but most are still a work in progress. Tellingly, discussions about the IoT tend to be focused on its potential for transformation rather than the here and now. That said, there are already important examples of where it is already playing an important role, such as in the energy sector, car industry, insurance and agriculture, but in most organisations the IoT is still seen as something that's looming on the horizon.

In classic "Next Big Thing" style, the IoT is a catch-all phrase that vendors are scrambling to make their own (see also cloud and big data, with which, incidentally, there are big areas of overlap). So the fact that many people are confused about the technology itself and its wider implications should not come as a surprise.

There is no doubt that it is on its way, however, and that few will remain untouched. Forward-looking organisations are already taking it into account in their strategic thinking, but many aspects of IoT are very new and rapid change makes it hard to draw up any sort of practical roadmap.

Computing's ongoing research into the subject, which will be unveiled at the forthcoming Internet of Things Business Summit in May, reveals a whole spectrum of opinion on the IoT, ranging from "there's nothing new under the sun" to "the industrial revolution 4.0". With this in mind we thought it would be a good idea to take a stroll around the different facets the IoT, and to summarise the current state of play, looking at how physical constraints in power, dimension and data transmission and the need for security are being addressed and what sort of new realities might emerge.

Quick links
The IoT? It's all about power A matter of protocol More protocols The hardware The cloud Security and privacy Use cases and examples Small things in big numbers / further reading

Want to know how IoT can help your enterprise? Come to Computing's Internet of Things Business Summit 2016 - it's free for qualifying end users

The Internet of Things: what, who, how and why

The IoT is poised to change everything. Here are the facts

It's all about power

When the internet was first conceived the power consumption of its infrastructure was an afterthought at best. Maximising processing grunt, interoperability and throughput was the name of the game, and if things got hot, well you just stuck in a big fan.

Deploying billions of tiny wireless devices powered by even tinier batteries changes this equation completely - for one thing, no one wants the job of changing the batteries of a million sensors. Instead of fading after a day like a smartphone, IoT batteries need to last for months or even years on a single charge, and where possible should be able to top themselves up by harvesting energy from the sun or ambient electromagnetic fields.

As well as being energy efficient, the devices themselves need to be simple and cheap to manufacture. In labs all over the world, microelectronics engineers are tweaking logic gates to minimise leakage, adjusting voltages to the bare minimum to reduce consumption, and squeezing new functionality into existing designs. And of course, battery technology is one of the hot areas of research.

Internet traffic is governed by a number of protocols: the transport layer has UDP and TCP; the internet layer IPv4 and IPv6; the link layer Ethernet and Wi-Fi; the application layer HTTP and DNS, and so on.These have been developed over the years to optimise file transfer between servers, PCs and other devices with a similar architecture. The devices that form the basis of the IoT look very different.

Traffic between IoT devices themselves and between those devices and the internet is governed by a whole new set of protocols. Since there is a direct trade-off between data throughput and energy efficiency, IoT protocols are heavily constrained versions of their internet counterparts, designed to extract the maximum value out of every packet of data and to only transmit it as far as it needs to go.

The Internet of Things: what, who, how and why

The IoT is poised to change everything. Here are the facts

A matter of protocol

The main tasks performed by IoT devices can be grouped into three areas: sensing, processing and wireless communication. A huge amount of work is being done in all of these areas by competing consortia of industry giants and government bodies, with each intent on creating the next VHS and avoiding backing the next Betamax. Competition in the area of protocols is fierce.

ZigBee and Bluetooth Low Energy

Where the internet is commonly accessed over wireless LAN, with the IoT the shorter range personal area network (PAN) predominates; while the internet is designed for file transfer, the IoT is optimised for communicating state, sending small packets of data in short bursts, for example to relay a temperature reading every few seconds.

Two of the most familiar names in the IoT world are the PAN technologies ZigBee and Bluetooth Low Energy (BLE). ZigBee is a wireless communication standard that was developed a decade ago explicitly to support the emerging IoT by providing low-power mesh networking to consumer devices. BLE, on the other hand, is a cut-down version of the popular Bluetooth standard that was created to compete with ZigBee.

Both have a range of a few meters, low latencies, and support extremely low-cost chipsets that consume little power. A typical BLE device can operate for a year or more on a small coin cell battery.

The biggest practical difference between them is architectural, namely the number of devices that can be connected together. BLE allows a maximum of seven slave devices to be controlled by one master, although the resulting star network may be joined with others to create a larger "scatternet". ZigBee allows up to 65,000 devices to be connected in a mesh and was designed with larger networks in mind.

In terms of use cases, ZigBee and BLE (and competitor Z-Wave) are pretty similar, being suited to sensor networks, smart buildings and small-scale industrial uses. Both are designed to sidestep interference from Wi-Fi and other signals operating in the same frequency ranges, although they do this in different ways.

Sigfox and LoRa

Where the internet enables long-range communication over wireless WANs, the IoT equivalent is the low-power wireless WAN or LPWAN. There are plenty of competing protocols here too, each with its own niche.

Sigfox, backed by Intel, is a low-complexity LPWAN protocol used for infrequent communication - a sensor that wakes up when a tank needs refilling perhaps, or one that checks whether a parking space is empty. Because of its simple needs the data rate is less than a kilobit per a second, but its operational range may be hundreds of miles.

LoRa allows a higher bit rate and operates over a wider range of frequencies. It is backed by IBM, Cisco and other members of the LoRa Alliance. As with ZigBee and BLE, while there are important technical differences between them, LoRa and Sigfox are sufficiently similar to be used for many of the same sorts of applications. And yes, there are others too: step forward nWave Weightless, Helium and RPMA.

The overlap in functionality between competing protocols is one factor that makes deciding on the best IoT strategy rather difficult.

The Internet of Things: what, who, how and why

The IoT is poised to change everything. Here are the facts

Yet more protocols

Apologies for the plethora of protocols but, in truth, protocols are what expanding the reach of the internet to encompass cheap, battery powered low-energy devices is all about. In any discussion of the IoT they are very hard to avoid.

The next set is about connecting the sea of sensors to the internet proper, which is where many devices such as home smart meters and smart energy grids live, as well as the dashboards and other user interfaces.

Traffic originating from the internet (IPV4 or IPv6) cannot be translated directly to the micro world of the IoT. One reason is that the minimum packet size for IPv6 is 1,280 bytes, whereas the IEEE 18.15.4 low-powered wireless protocol used by many IoT "things" stipulates a maximum packet size of 127.

Some protocols (e.g. ZigBee) get around this size mismatch using gateway devices, but the snappily named 6LoWPAN (IPv6 over Low power Wireless Personal Area Networks) does so by compressing the IPv6 addresses so that IPv6 packets can be carried on wireless IEEE 18.15.4 networks. Because it allows a simpler interface with a wider range of IP networks including Ethernet and Wi-Fi, some people think 6LoWPAN will take a bite out of ZigBee's share of the home-user market as standard ZigBee devices require a more complex gateway setup (although ZigBee has now hit back with an IP version of its protocol).

CoAP and MQTT

The web applications layer of the internet has its IoT equivalent too. CoAP (Constrained Application Protocol) is the machine-to-machine version of HTTP designed to meet the needs of embedded systems where memory, storage and processing power are all in short supply and packets of data are small. Via a REST interface (the API specification widely used in web applications) IoT devices can perform familiar functions like SEND, GET, DELETE and so on. So CoAP is used to control switches, valves and sensors over the internet.

For efficiency and integration with a wide range of devices, both CoAP and 6LoWPAN use UDP as the transport layer. But UDP is an unreliable protocol, which limits their use to tasks that are not sensitive to latency. Using them to coordinate traffic lights, for example, would soon end up with a nasty accident.

The messaging protocol MQTT runs over the reliable TCP transport layer. MQTT is designed for sensor networks in remote locations as a way to transmit data from sensors to servers, for early warning systems perhaps, or sensors on a dam or pipeline, or to track the vital signs of patients after they have left hospital.

Another messaging protocol is XMPP, which offers more chat functionality but with a larger overhead as it requires a persistent TCP connection.

To some in the industry, the proliferation of standards and protocols is a worrisome indication of the immaturity of the IoT.

"One of the issues with the IoT is that there are no standards," said Paul Scholey of integration company Pentaho. "So, different devices from different manufacturers that even aspire to do the same things can have radically different data formats and potentially different touch-points."

The Internet of Things: what, who, how and why

The IoT is poised to change everything. Here are the facts

The hardware

IoT "things" comprise sensors, actuators, transmitters, routers and controllers. All must be compact, simple, cheap to produce, robust, energy efficient, and able to connect to other devices as required.

Rather than constantly creating new designs, it is often cheaper to add functionality to existing chipsets to produce multi-modal devices - so a light sensor may be adapted to sense temperature and humidity too. Another popular sensor is the initial measurement unit (IMU) that measures velocity, orientation and gravitational force. The IMU in an iPhone costs only a few pounds. Other sensors worth a mention are LiDAR modules, which bounce laser light off targets to build 3D images, and ultrasonic rangers that measure distance.

Examples of actuators include switches, lights, alarms and motors, which may be connected to robots, air conditioning units, cameras and other machinery.

At the small end of the IoT hardware scale are the single-purpose sensors; at the larger end are general-purpose machines like the Raspberry Pi and TI BeagleBoard.

Unlike their mains-connected internet counterparts, IoT sensors typically switch on and off several times a second to conserve energy. Some even sleep indefinitely until they are sent a reset signal. Many devices are a simple system on a chip (SoC) with radio frequency functions, some computing power and a little bit of RAM. They may run a cut-down embedded version Linux designed to support a specific set of tasks. The Raspberry Pi can run general-purpose Linux to support a wide variety of functions.

Aside from their functionality, the architecture of these things depends largely on their communication needs. For example, if they are just pushing small packets of data to and fro between themselves in a sensor network, their requirement for bandwidth will be less than if they are connected directly to the larger internet or have to carry video or other streamed data.

Moving on to the vendors of these components, it is hard to find an IT manufacturer that is not riding the IoT wave at present, or at least attempting to, but key players include ARM, which has pioneered low energy chip design, Intel, Cisco, Qualcomm, Texas Instruments, Ericsson, GE, Freescale, Amtel and Bosch.

The Internet of Things: what, who, how and why

The IoT is poised to change everything. Here are the facts

The cloud

It is perfectly possible to have a smart home whose appliances and lighting systems only speak to one another (as well as their owner), just as most new cars have entertainment systems that connect to headsets and smartphones but not to the outside world - but that would be kind of missing the point. The smart home that can receive localised weather forecasts and automatically adjust its climate control systems to suit, and the car that avoids traffic jams or pollution hot-spots by networking with other vehicles, and whose windscreen wipers are connected to weather databases allowing pinpoint mapping of wet conditions, are staple IoT visions. To fulfil their potential cloud is required to provide both the overview and the necessary service integration. Large-scale multi-faceted IoT use cases, such as smart cities and energy grids, will be simply impossible to realise in the absence of a cloud-based co-ordination layer.

The relentless increase in both the number of sensors and the data produced makes scalable cloud platforms a good fit for IoT too. Cloud platforms can offer flexible storage and compute for processor-intensive analytics, feeding back information and taking the load off onboard systems, which in the case of IoT are often limited, particularly for real-time applications.

Pretty much all the major cloud vendors now offer special IoT cloud services - although most are pretty new and no doubt a certain amount of rebadging has taken place.

Most of the broader IoT platforms that have emerged or are in development also have a cloud element. Many integrate connectivity with analytics: IBM's Watson IoT Platform extends cognitive computing services to the IoT.

With a focus on the industrial IoT, GE's Predix is a platform that connects machines including wind farms and jet engines with big data analytics for real-time control and which allows organisations to build their own applications on the platform.

There are numerous other IoT platforms too, from integrated collections aimed at hobbyists and small home networks like Apple's HomeKit, to giants like Amazon, whose AWS IoT is a managed cloud platform that lets connected devices interact with other cloud-based applications and devices, and Microsoft's Azure Hub - a similar end-to-end IoT platform.

It should be repeated, however, that most of these platforms are very new - AWS IoT is only just out of beta - and there is still much work to be done around standards and security.

The Internet of Things: what, who, how and why

The IoT is poised to change everything. Here are the facts

Security and privacy

"Then those Things ran about with big bumps, jumps and kicks and with hops and big thumps and all kind of tricks. And I said, 'I do not like the way that they play! If Mother could see this, oh, what would she say?'" Dr Seuss, The Cat in the Hat

What happens if the things run amok? It is widely acknowledged that security of the internet and Wi-Fi is poor precisely because it has been bolted on as an afterthought rather than being built in from the start, and many fear the same is about to happen with IoT. And the consequences, they point out, could be even more serious.

Baby monitors, connected coffee machines and Jeeps have already been the subject of recent high-profile hacks. Bearing in mind that sensors and actuators are soon going to be everywhere - in our homes, our cars, in chemical plants, dams and nuclear power stations - the implications of an insecure IoT look distinctly dystopian.

One of the big issues with the heterogenous, multi-vendor, multi-protocol, free-for-all gold rush that characterises the IoT at present is that it creates many overlooked nooks and crannies, vulnerabilities that potentially present a huge attack surface for anyone who might want to compromise it.

An attacker may be able to spoof an access point and redirect traffic, and any flaw in the complex setup of inter-device communications might provide an entry point. Granted, most of the communications protocols mentioned above have encryption built in, and some allow for key management, certificates and other security procedures, but not all protections are switched on by default.

Then there is the issue of updates. The middleware run by the multitude of devices must obviously be patched automatically over the network, but what if that network goes down, or if a device fails to connect to it? What if the component vendor goes out of business?

And then there is the privacy angle. Many people now use fitness trackers and other wearables to monitor their activity, and very similar technology is deployed in the healthcare sector. What happens to that data when it is beamed up to the cloud? How is it stored? Who can access it? How is it encrypted? Will it be sold off to insurance companies who might use it to bump up policy prices? Will regulations be able to keep up with technology in a way they haven't in the past?

These are not new issues, of course, but the huge increase in connectivity, the types of devices that are connected and the vast quantities of highly sensitive data that will be "out there" mean that privacy and security by design must be priorities rather than afterthoughts this time around. Otherwise the consequences of failure could be very severe indeed.

The Internet of Things: what, who, how and why

The IoT is poised to change everything. Here are the facts

Use cases and examples

On now to the current applications and potential use cases of the IoT. We've already mentioned smart homes and fitness bands, and indeed consumer applications have arrived more quickly than business ones, although industrial control systems deploying something that looks very like IoT have been around for years.

Many IoT use cases are associated with the field of machine learning, and more broadly, big data. Indeed from an analytcs point of view, the IoT may be seen as big data on steroids.

Cognitive computing

Cognitive computing describes machines that learn through experience and are capable of adapting themselves to solve new problems without human assistance. It is a technology that's still in the early stages of development, but arguably practical implementations are already on the market in the shape of smart meters like Google's Nest, which learn users' preferred temperature at certain times of the day or months of the year and adjust heating and air conditioning accordingly. The longer it is installed, the more it leans.

Neural networks

Neural networks are modelled on the human brain and use large quantities of data to solve problems such as speech recognition and natural language processing, learning and fine tuning processes as they go.

Robotics

Today's robots are mostly designed to do repetitive jobs in a predictable environment, but armed with sensors and cameras that can take thousands of photos of an object from various angles and in different light conditions, robots can be trained to recognise objects and ultimately taught how to handle them. The holy grail is to have robots that can operate in a home or a hospital or supermarket and interact directly (and predictably) with humans. Such robots would need to be able to cope with variability of environment and unexpected events, and this requires computing power and storage that exceed on-board capacity, which implies the cloud.

Self-driving vehicles

Arguably an early realisation of such a robot is the self-driving car. These vehicles need to know where they are, where they are going, and how to get there, and they must learn to react instantly and appropriately to a huge variety of scenarios - a pedestrian stepping into the road, overtaking a cyclist, pulling out of a junction. They must be able to recognise familiar objects in unfamiliar settings and light conditions (something that is intuitive for humans but very hard for robots) with the consequences of any error being potentially fatal.

Telemetrics

For a few years now insurance companies have offered lower priced policies to those that drive fewer than a certain number of miles per year. The quid pro quo, of course, is that your driving is tracked and recorded. This type of individualised insurance through surveillance is likely to become more common, moving to health and household insurance.

Smart contracts

What if your washing machine could order its own detergent when it gets low? Welcome to the world of smart contracts, where machines contract with other machines without intervention from people. Experimental work is going on with smart contracts using blockchain technology in finance, insurance, legal and future homes.

Smart cities

The city of Santander in Spain has many thousands of sensors collecting data on everything from parking space availability to air quality. It is one of many metropolitan areas around the world that are experimenting with the smart city concept. In other places traffic lights automatically adjust their phases to provide maximum traffic flow, energy consumption is being managed by demand-response and smart grid technologies, plants are watered when soil sensors detect that they are dry, and parking and public transport are priced dynamically.

Utilities

Water and oil companies have been using the IoT-type technologies for years, with sensors and data loggers monitoring flows through pipes and water levels in tanks and rivers. But the rise of real-time analytics now means they can do this in a more predictive fashion, fixing machinery and changing settings before issues even become obvious. In the energy sector, the IoT is vital in optimising wind farms where the conditions may change many times per second.

Agriculture

Many crops are acutely sensitive to local conditions. Sensors are being used to optimise these conditions on a very localised basis by controlling the application of water, pesticides and herbicides.

Sport

Sports from tennis to motor racing are attaching sensors to cars, arms and legs to achieve that all important one percent advantage over rivals.

The Internet of Things: what, who, how and why

The IoT is poised to change everything. Here are the facts

Small things in big numbers

No overview of the IoT would be complete without a selection of big numbers, so here are a few we dug up.

6.4 billion connected things will be in use by the end of 2016, estimates analyst firm Gartner. This represents a 30 per cent increase from 2015. These numbers mean that more than five million things are getting connected every day. By 2020 the number of connected things will be 21 billion.

McKinsey & Co predicts that the economic impact of the IoT could be between US$3.9tn and US$11.1tn per year by 2025.

Who knows whether these projections will turn out to be accurate. What is certainly true though, is that no business will remain untouched by the IoT. It is already here and the direction of travel is clear, as the following articles from Computing demonstrate.

Further reading