IoT and Edge Computing for Architects - Second Edition

By Perry Lea
  • Instant online access to over 7,500+ books and videos
  • Constantly updated with 100+ new titles each month
  • Breadth and depth in over 1,000+ technologies
  1. IoT and Edge Computing Definition and Use Cases

About this book

Industries are embracing IoT technologies to improve operational expenses, product life, and people's well-being. An architectural guide is needed if you want to traverse the spectrum of technologies needed to build a successful IoT system, whether that's a single device or millions of IoT devices.

IoT and Edge Computing for Architects, Second Edition encompasses the entire spectrum of IoT solutions, from IoT sensors to the cloud. It examines modern sensor systems, focusing on their power and functionality. It also looks at communication theory, paying close attention to near-range PAN, including the new Bluetooth® 5.0 specification and mesh networks. Then, the book explores IP-based communication in LAN and WAN, including 802.11ah, 5G LTE cellular, Sigfox, and LoRaWAN. It also explains edge computing, routing and gateways, and their role in fog computing, as well as the messaging protocols of MQTT 5.0 and CoAP.

With the data now in internet form, you'll get an understanding of cloud and fog architectures, including the OpenFog standards. The book wraps up the analytics portion with the application of statistical analysis, complex event processing, and deep learning models. The book then concludes by providing a holistic view of IoT security, cryptography, and shell security in addition to software-defined perimeters and blockchains.

Publication date:
March 2020
Publisher
Packt
Pages
632
ISBN
9781839214806

 

IoT and Edge Computing Definition and Use Cases

You wake up Tuesday, May 17, 2022, around 6:30 A.M. PST, as you always do. You never really needed an alarm clock. You are one of those types with some form of physiological clock. Your eyes open to a fantastic sunny morning as it's approaching 70°F outside. You will take part in a day that will be completely different than the morning of Wednesday, May 17, 2017. Everything about your day, your lifestyle, your health, your finances, your work, your commute, even your parking spot will be different. Everything about the world you live in will be different: energy, healthcare, farming, manufacturing, logistics, mass transit, environment, security, shopping, and even clothing. This is the impact of connecting ordinary objects to the Internet, or the Internet of Things (IoT). I think a better analogy is the Internet of Everything.

Before you even awakened, a lot has happened in the IoT that surrounds you. Your sleep behavior has been monitored by a sleep sensor or smart pillow. Data was sent to an IoT gateway and then streamed to a cloud service you use for free that reports to a dashboard on your phone. You don't need an alarm clock, but if you had a 5 A.M. flight, you would set it—again, controlled by a cloud agent using the if this, then that (IFTTT) protocol. Your dual-zone furnace is connected to a different cloud provider and is on your home 802.11 Wi-Fi, as are your smoke alarms, doorbell, irrigation systems, garage door, surveillance cameras, and security system. Your dog is chipped with a proximity sensor using an energy harvesting source that lets him open the doggy door and tell you where he is.

You don't really have a PC anymore. You certainly have a tablet-style computer and a smartphone as your central creation device, but your world is based on using a VR/AR headset since the screen is so much better and larger. You do have an edge computing gateway in your closet. It's connected to a 5G service provider to get you on the Internet and WAN because wired connections don't work for your lifestyle—you are mobile, connected, and online no matter where you are, and 5G and your favorite carrier make sure your experience is great in a hotel room in Miami or your home in Boise, Idaho. The gateway also performs a lot of actions in your home for you, such as processing video streams from those webcams to detect whether there's been a fall or an accident in the house. The security system is being scanned for anomalies (strange noises, possible water leaks, lights being left on, your dog chewing on the furniture again). The edge node also acts as your home hub, backing up your phone daily because you have a tendency to break them, and serves as your private cloud even though you know nothing about cloud services.

You ride your bike to the office. Your bike jersey uses printable sensors and monitors your heart rate and temperature. That data is streamed over Bluetooth Low Energy to your smartphone simultaneously while you listen to Bluetooth audio streamed from your phone to your Bluetooth earphones. On the way there, you pass several billboards all displaying video and real-time ads. You stop at your local coffee shop, and there is a digital signage display out front calling you out by name and asking if you want the last thing you ordered yesterday: a 12 oz Americano with room for cream. It did this by a beacon and gateway recognizing your presence within five feet and approaching the display. You select yes, of course. Most people arrive at work via their car and are directed to the optimal parking space via smart sensors in each parking slot. You, of course, get the optimal parking space right out front with the rest of the cyclists.

Your office is part of a green energy program. Corporate policies mandate a zero-emission office space. Each room has proximity sensors to detect not only whether a room is occupied, but also who is in the room. Your name badge to get in the office is a beaconing device on a 10-year battery. Your presence is known once you enter the door. Lights, HVAC, automated shades, ceiling fans, even digital signage are connected. A central fog node monitors all the building information and syncs it to a cloud host. A rules engine has been implemented to make real-time decisions based on occupancy, time of day, and the season of the year, as well as inside and outside temperatures. Environmental conditions are ramped up or down to maximize energy utilization. There are sensors on the main breakers listening to the patterns of energy and making a decision on the fog nodes if there are strange patterns of energy usage that need examination.

It does all this with several real-time streaming edge analytics and machine learning algorithms that have been trained on the cloud and pushed to the edge.

The office hosts a 5G small cell to communicate externally to the upstream carrier, but it also hosts a number of small-cell gateways internally to focus signals within the confines of the building. The internal 5G acts as a LAN as well.

Your phone and tablet have switched to the internal 5G signal, and you switch on your software-defined network overlay and are instantly on the corporate LAN. Your smartphone does a lot of work for you; it is essentially your personal gateway to your own personal area network surrounding your body. You drop into your first meeting today, but your co-worker isn't there and arrives a few minutes late. He apologizes but explains his drive to work was eventful.

His newer car informed the manufacturer of a pattern of anomalies in the compressor and turbocharger. The manufacturer was immediately informed of this, and a representative called your co-worker to inform him that the vehicle has a 70 percent chance of having a failed turbo within two days of his typical commute. They scheduled an appointment with the dealership and have the new parts arriving to fix the compressor. This saved him considerable cost in replacing the turbo and a lot of aggravation.

For lunch, the team decides to go out to a new fish taco place downtown. A group of four of you manage your way into a coupe more comfortable for two and make your way. Unfortunately, you'll have to park in one of the more expensive parking structures.

Parking rates are dynamic and follow a supply-and-demand basis. Because of some events and how full the lots are, the rates doubled even for midday Tuesday. On the bright side, the same systems raising the parking fees also inform your car and smartphone exactly which lots and which space to drive to. You punch in the fish taco address, the lot and capacity pop up, and you reserve a spot before you arrive. The car approaches the gate, which identifies your phone signature, license plate, or a combination of multiple factors and opens up. You drive to the spot, and the application registers with the parking cloud that you are in the right spot over the correct sensor.

That afternoon, you need to go to the manufacturing site on the other side of town. It's a typical factory environment: several injection molding machines, pick-and-place devices, packaging machines, and all the supporting infrastructure. Recently, the quality of the product has been slipping. The final product has joint connection problems and is cosmetically inferior to last month's lot. After arriving at the site, you talk to the manager and inspect the site. Everything appears normal, but the quality certainly has been marginalized. The two of you meet and bring up the dashboards of the factory floor.

The system uses a number of sensors (vibration, temperature, speed, vision, and tracking beacons) to monitor the floor. The data is accumulated and visualized in real time. There are a number of predictive maintenance algorithms watching the various devices for signs of wear and error. That information is streamed to the equipment manufacturer and your team as well. The manufacturing automation and diagnostics logs didn't pick up any abnormal patterns, as they had been trained by your best experts. This looks like the type of problem that would turn hours into weeks and force the best and brightest in your organization to attend expensive daily SWOT (strengths, weaknesses, opportunities, and threats) team meetings. However, you have a lot of data. All the data from the factory floor is preserved in a long-term storage database. There was a cost to that service. At first, the cost was difficult to justify, but now you believe it may have paid for itself a thousand-fold. Taking all that historical data through a complex event processor and analytics package, you quickly develop a set of rules that model the quality of your failing parts. Working backward to the events that led to the failures, you realize it is not a point failure, but has several aspects:

  • The internal temperature of the working space rose 2°C to conserve energy for the summer months.
  • The assembly slowed down output by 1.5 percent due to supply issues.
  • One of the molding machines was nearing a predictive maintenance period, and the temperature and assembly speed pushed its failing case over the predicted value.

You found the issue and retrained the predictive maintenance models with the new parameters to catch this case in the future. Overall, not a bad day at work.

While this fictional case may or may not be true, it's pretty close to reality today. Wikipedia defines the IoT this way: The Internet of things (IoT) is the inter-networking of physical devices, vehicles (also referred to as "connected devices" and "smart devices"), buildings, and other items embedded with electronics, software, sensors, actuators, and network connectivity which enable these objects to collect and exchange data. (https://en.wikipedia.org/wiki/internet_of_things)

 

History of the IoT

The term "IoT" can most likely be attributed to Kevin Ashton in 1997 and his work at Procter and Gamble using RFID tags to manage supply chains. The work brought him to MIT in 1999 where he and a group of like-minded individuals started the Auto-ID Center research consortium (for more information, visit http://www.smithsonianmag.com/innovation/kevin-ashton-describes-the-internet-of-things-180953749/).

Since then, IoT has taken off from simple RFID tags to an ecosystem and industry that will have 1 trillion Internet-connected devices by 2030. The concept of things being connected to the Internet up through 2012 was primarily connected smartphones, tablets, PCs, and laptops. Essentially, things that first functioned in all respects as a computer. Since the humble beginnings of the Internet, starting with ARPANET in 1969, most of the technologies surrounding the IoT didn't exist. Up to the year 2000, most devices that were associated with the Internet were, as stated, computers of various sizes. The following timeline shows the slow progress in connecting things to the Internet:

Year Device Reference

1973

Mario W. Cardullo receives the patent for first RFID tag.

US Patent US 3713148 A

1982

Carnegie Mellon Internet-connected soda machine.

https://www.cs.cmu.edu/~coke/history_long.txt

1989

Internet-connected toaster at Interop '89.

IEEE Consumer Electronics Magazine (Volume: 6, Issue: 1, Jan. 2017)

1991

HP introduces HP LaserJet IIISi: the first Ethernet-connected network printer.

http://hpmuseum.net/display_item.php?hw=350

1993

Internet-connected coffee pot at University of Cambridge (the first Internet-connected camera).

https://www.cl.cam.ac.uk/coffee/qsf/coffee.html

1996

General Motors OnStar (2001 remote diagnostics).

https://en.wikipedia.org/wiki/OnStar

1998

Bluetooth Special Interest Group (SIG) formed.

https://www.bluetooth.com/about-us/our-history

1999

LG Internet Digital DIOS refrigerator.

https://www.telecompaper.com/news/lg-unveils-internetready-refrigerator--221266

2000

First instances of the Cooltown concept of pervasive computing everywhere: HP Labs, a system of computing and communication technologies that, combined, create a web-connected experience for people, places, and objects.

https://www.youtube.com/watch?v=U2AkkuIVV-I

2001

First Bluetooth product launched: KDDI Bluetooth-enabled mobile phone.

http://edition.cnn.com/2001/BUSINESS/asia/04/17/tokyo.kddibluetooth/index.html

2005

United Nation's International Telecommunications Union report predicting the rise of IoT for the first time.

http://www.itu.int/osg/spu/publications/internetofthings/internetofThings_summary.pdf

2008

IPSO Alliance formed to promote IP on objects, first IoT-focused alliance.

https://www.ipso-alliance.org

2010

The concept of Smart Lighting formed after success in developing solid-state LED light bulbs.

https://www.bu.edu/smartlighting/files/2010/01/BobK.pdf

2014

Apple creates iBeacon protocol for beacons.

https://support.apple.com/en-us/HT202880

Certainly, the term IoT has generated a lot of interest and hype. One can easily see that from a buzzword standpoint. The number of patents issued (https://www.uspto.gov) has grown exponentially since 2010. The number of Google searches (https://trends.google.com/trends/) and IEEE peer-reviewed paper publications hit the knee of the curve in 2013:

Figure 1: Analysis of keyword searches for IoT, patents, and technical publications

 

IoT potential

The IoT is already affecting every segment in industrial, enterprise, health, and consumer products. It is important to understand the impact, as well as why these disparate industries will be forced to change in the ways they build products and provide services. Perhaps your role as an architect forces you to focus on one particular segment; however, it is helpful to understand the overlap with other use cases.

As previously mentioned, there is an opinion that the impact of IoT-related industries, services, and trade will affect 3 percent (The route to a trillion devices, ARM Ltd 2017) to 4 percent (The Internet of Things: Mapping Value Beyond the Hype, McKinsey and Company 2015) of global GDP by 2020 (extrapolated). Global GDP for 2016 was $75.64 trillion dollars, with an estimate that by 2020 it will rise to $81.5 trillion. That provides a range of value from IoT solutions of $2.4 trillion to about $4.9 trillion.

The scale of connected objects is unprecedented. Speculation of industry growth is imperiled with risks. To help normalize the impact, we look at several research firms and reports on the number of connected objects. The range is large, but still in the same order of magnitude. The average of these 10 analyst forecasts is about 33.4 billion connected things by 2020-2021. ARM recently conducted a study and forecast that by 2035 one trillion connected devices will be operational. By all accounts, the IoT deployment growth rate in the near term is about 20 percent year over year.

Figure 2: Analysts and industry claims of the number of connected objects

These numbers should at first glance impress the reader. For example, if we took a very conservative stance and predicted that only 20 billion newly connected devices would be deployed (excluding the traditional computing and mobile products), we would be saying that 211 new Internet-connected objects will come online every second.

Why this is of significance to the technology industry and IT sector is the fact that world population currently has a growth rate of roughly 0.9 percent to 1.09 percent per year (https://esa.un.org/unpd/wpp/). World population growth rate peaked in 1962 at 2.6 percent year over year and has steadily been declining due to a number of factors. First and foremost, improvement in world GDP and economies has a propensity to reduce birth rates. Other factors include wars and famine. That growth implies that human-connected objects will plateau, and machine to machine (M2M) and connected objects will represent the majority of devices connected to the Internet. This is important because the IT industry applies value to a network not necessarily by how much data is consumed, but by how many connections there are. This, generally speaking, is Metcalfe's law, and we will talk about that later in this book. It is also worth noting that after the first public website went live at CERN in 1990, it then took 15 additional years for 1 billion people to be regular users of the Internet. IoT is looking to add 6 billion connected devices per year. This, of course, is swaying the industry.

Figure 3: The disparity between human population growth versus connected thing growth. The trend has been a 20 percent growth of connected objects versus a nearly flat 0.9 percent human growth. Humans will no longer drive network and IT capacity.

It should be noted that economic impact is not solely revenue generation. The impact from IoT or any technology comes in the form of:

  • New revenue streams (for example, green energy solutions)
  • Reducing costs (for example, in-home patient healthcare)
  • Reducing time to market (for example, factory automation)
  • Improving supply chain logistics (for example, asset tracking)
  • Reducing production loss (for example, theft or spoilage of perishables)
  • Increasing productivity (for example, machine learning and data analytics)
  • Cannibalization (for example, Nest replacing traditional thermostats)

In our discussion throughout this book, it should be at the top of our minds as to what value an IoT solution delivers. If it is simply a new gadget, there will be a limited market scope. Only when the foreseeable benefit outweighs the cost will an industry thrive.

Generally speaking, the target used should be a 5x improvement over a traditional technology. That has been my goal in the IT industry. When considering the cost of change, training, acquisition, support, and so on, a 5x differential is a fair rule of thumb.

 

Definition of the Internet of Things

One should look at some of these claims with a degree of skepticism. It is nearly impossible to quantify the exact number of devices that are Internet-connected. Additionally, we have to separate those devices that are naturally Internet-connected like smartphones, PCs, servers, network routers, and IT infrastructure. We should also not include in the realm of IoT those machines that have had presence in offices, homes and workplaces for decades that are essentially connected through some form of networking. We do not include office printers, copiers, or scanners as part of the IoT spectrum.

This book will examine IoT from the perspective of connecting devices that have not necessarily been connected to each other or the Internet. These devices may have historically not had much if any computational or communication abilities. By that, we imply that devices historically have had cost, power, space, weight, size, or thermal limits.

As we see in the history of IoT devices, connecting traditionally unconnectable objects like refrigerators at Carnegie Mellon has been possible since the early 1980s, but the cost was significant. It required the processing power of a DEC PDP11 mainframe computer. Moore's Law demonstrates the increases in the number and density of transistors in silicon chipsets, while Dennard scaling improves the power profile of computers. With these two trends, we now produce devices that utilize more powerful CPUs and increased memory capacity and run operating systems capable of executing a full network stack. It is only with these requirements being met that the IoT has become an industry unto itself.

The basic requirements of a device to be considered part of the IoT:

  • Computationally capable of hosting an Internet protocol software stack
  • Hardware and power capable of utilizing a network transport such as 802.3
  • Not a traditional Internet-connected device, such as a PC, laptop, smartphone, server, data center appliance, office productivity machine, or tablet computer

We also include "edge" devices in this book. Edge devices themselves can be IoT devices or can "host" IoT devices. Edge devices as detailed later in this book will generally be managed computer nodes that extend closer to the sources of data generation or data action. They may not be typical servers and clusters found in data centers but space, power, and environmentally hardened devices that are in the field. For example, a data center blade would consist of electronics optimized for the climate-controlled atmosphere of a server farm with hot and cold aisles, heat exchangers, and uninterruptible power supplies. Edge devices may be found outside and exposed to weather elements and in areas where constant and consistent power is not an option. Other times, they may include traditional server nodes, but outside the constraints of a datacenter.

Given these qualifiers, the true size of the IoT market is smaller than analyst forecasts. When we divide traditional IT and Internet-connected devices from IoT devices, we see a different growth rate as shown in the following figure.

Figure 4: Separating sales volume of IoT devices by definition from non-IoT devices (for example, IT equipment and mobile computing)

Further analysis into actual components that are used in IoT devices reveals another interesting pattern. As already mentioned, most Internet-connected devices require a certain level of performance and hardware to communicate through standard protocols. Yet the following graphic shows a difference in the number of communication chips and processors versus the number of sensors that are shipping. This reinforces the concept that there is a large fan-out from sensors to edge computers and communication devices.

Figure 5: Trend in sales of sensors, processors, and communication ICs within IoT sales

What is notable is that most IoT installations are not a single device that has the capabilities of running an Internet hardware and software stack. Most sensors and devices have no capabilities of reaching the Internet directly. They lack the processing capabilities, memory resources, and power distribution required for full Internet connectivity. Rather, much of what is really the IoT relies upon gateways and edge computers in a hub-and-spoke model. There is a large fan-out of devices that connect to edge computers through local personal area networks, non-IP networks (Bluetooth), industrial protocols (ModBus), legacy brownfield protocols (RS232), and hardware signals.

Industry and manufacturing

Industrial IoT (IIoT) is one of the fastest-growing and largest segments in the overall IoT space by the number of connected things and the value those services bring to manufacturing and factory automation. This segment has traditionally been the world of operations technology (OT). This involves hardware and software tools to monitor physical devices in real time. These systems historically have been on-premises computers and servers to manage factory floor performance and output. We call these systems supervisory control and data acquisition (SCADA). Traditional information technology roles have been administered differently than OT roles. OT will be concerned with yield metrics, uptime, real-time data collection and response, and systems safety. The IT role will concentrate on security, groupings, data delivery, and services. As the IoT becomes prevalent in industry and manufacturing, these worlds will combine especially with predictive maintenance from thousands of factory and production machines to deliver an unprecedented amount of data to private and public cloud infrastructure.

Some of the characteristics of this segment include the need to provide near real-time or actual real-time decisions for OT. This means latency is a major issue for IoT on a factory floor.

Additionally, downtime and security are top concerns. This implies the need for redundancy and possibly private cloud networks and data storage. The industrial segment is one of the fastest-growing markets. One nuance of this industry is the reliance on brownfield technology, meaning hardware and software interfaces that are not mainstream. It is often the case that 30-year-old production machines rely on RS485 serial interfaces rather than modern wireless mesh fabrics.

Industrial and manufacturing IoT use cases

Following are the industrial and manufacturing IoT use cases and their impact:

  • Preventative maintenance on new and preexisting factory machinery
  • Throughput increase through real-time demand
  • Energy savings
  • Safety systems such as thermal sensing, pressure sensing, and gas leaks
  • Factory floor expert systems

Consumer

Consumer-based devices were one of the first segments to adopt things being connected to the Internet. Consumer IoT first took the form of a connected coffee pot at a university in the 1990s. It flourished with the adoption of Bluetooth for consumer use in the early 2000s.

Now millions of homes have Nest thermostats, Hue lightbulbs, Alexa assistants, and Roku set-top boxes. People too are connected with Fitbits and other wearable technology. The consumer market is usually the first to adopt these new technologies. We can also think of these as gadgets. All are neatly packaged and wrapped devices that are essentially plug and play.

One of the constraints in the consumer market is the bifurcation of standards. We see, for example, several WPAN protocols have a footing like Bluetooth, Zigbee, and Z-wave (all being non-interoperable).

This segment also has common traits with the healthcare market, which has wearable devices and home health monitors. We keep them separate for this discussion, and healthcare will grow beyond simple connected home health devices (for example, beyond the functionality of a Fitbit).

Consumer IoT use cases

The following are some of the consumer IoT use cases:

  • Smart home gadgetry: Smart irrigation, smart garage doors, smart locks, smart lights, smart thermostats, and smart security
  • Wearables: Health and movement trackers, smart clothing/wearables
  • Pets: Pet location systems, smart dog doors

Retail, finance, and marketing

This category refers to any space where consumer-based commerce transacts. This can be a brick-and-mortar store or a pop-up kiosk. These include traditional banking services and insurers, but also leisure and hospitality services. The retail IoT impact is already in process, with the goal of lowering sales costs and improving customer experience. This is done with a myriad of IoT tools. For simplicity in this book, we also add advertising and marketing to this category.

This segment measures value in immediate financial transactions. If the IoT solution is not providing that response, its investment must be scrutinized. This drives constraints on finding new ways to either save costs, or drive revenue. Allowing customers to be more efficient allows retailers and service industries to provide better customer experiences while minimizing overhead and loss in the cost of sales.

Retail, finance, and marketing IoT use cases

Some of the IoT use cases are as follows:

  • Targeted advertising, such as locating known or potential customers by proximity and providing sales information.
  • Beaconing, such as proximity sensing customers, traffic patterns, and inter-arrival times as marketing analytics.
  • Asset tracking, such as inventory control, loss control, and supply chain optimizations.
  • Cold storage monitoring, such as analyze cold storage of perishable inventory. Apply predictive analytics to food supply.
  • Insurance tracking of assets.
  • Insurance risk measurement of drivers.
  • Digital signage within retail, hospitality, or citywide.
  • Beaconing systems within entertainment venues, conferences, concerts, amusement parks, and museums.

Healthcare

The healthcare industry will contend with manufacturing and logistics for the top spot in revenue and impact on IoT. Any and all systems that improve the quality of life and reduce health costs are a top concern in nearly every developed country. The IoT is poised to allow for remote and flexible monitoring of patients wherever they may be.

Advanced analytics and machine learning tools will observe patients in order to diagnose illness and prescribe treatments. Such systems will also be the watchdogs in the event of needed life-critical care. Currently, there are about 500 million wearable health monitors, with double-digit growth in the years to come.

The constraints on healthcare systems are significant. From HIPAA compliance to the security of data, IoT systems need to act like hospital-quality tools and equipment. Field systems need to communicate with healthcare centers 24/7, reliably and with zero downtime if the patient is being monitored at home. Systems may need to be on a hospital network while monitoring a patient in an emergency vehicle.

Healthcare IoT use cases

Some of the healthcare IoT use cases are as follows:

  • In-home patient care
  • Learning models of predictive and preventative healthcare
  • Dementia and elderly care and tracking
  • Hospital equipment and supply asset tracking
  • Pharmaceutical tracking and security
  • Remote field medicine
  • Drug research
  • Patient fall indicators

Transportation and logistics

Transportation and logistics will be a significant, if not the leading driver in IoT. The use cases involve using devices to track assets being delivered, transported, or shipped, whether that's on a truck, train, plane, or boat. This is also the area of connected vehicles that communicate to offer assistance to the driver, or preventative maintenance on behalf of the driver. Right now, an average vehicle purchased new off a lot will have about 100 sensors. That number will double as vehicle-to-vehicle communication, vehicle-to-road communication, and automated driving become must-have features for safety or comfort. This has important roles beyond consumer vehicles and extends to rail lines and shipping fleets that cannot afford any downtime. We will also see service trucks that can track assets such as workers' tools, construction equipment, and other valuable assets. Some of the use cases can be very simple, but also very costly, such as monitoring the location of service vehicles in the delivery of stock.

Systems are needed to automatically route trucks and service personnel to locations based on demand versus routine.

This mobile-type category has the requirement of geolocation awareness. Much of this comes from GPS navigation. From an IoT perspective, the data analyzed would include assets and time, but also spatial coordinates.

Transportation and logistics IoT use cases

Following are some of the transportation and logistics IoT use cases:

  • Fleet tracking and location awareness
  • Municipal vehicle planning, routing and monitoring (snow removal, waste disposal)
  • Cold storage transportation and safety of food delivery
  • Railcar identification and tracking
  • Asset and package tracking within fleets
  • Preventative maintenance of vehicles on the road

Agricultural and environment

Farming and environmental IoT includes elements of livestock health, land and soil analysis, micro-climate predictions, efficient water usage, and even disaster predictions in the case of geological and weather-related disasters. Even as the world population growth slows, world economies are becoming more affluent. Even as famines are less common than 100 years ago, the demand for food production is set to double by 2035. Significant efficiencies in agriculture can be achieved through IoT. Using smart lighting to adjust the spectrum frequency based on poultry age can increase growth rates and decrease mortality rates based on stress on chicken farms. Additionally, smart lighting systems could save $1 billion annually on energy versus the common dumb incandescent lighting currently used. Other uses include detecting livestock health based on sensor movement and positioning. A cattle farm could find animals with the propensity of sickness before a bacterial or viral infection were to spread. Remote edge analysis systems could find, locate, and isolate heads of cattle in real time, using data analytics or machine learning approaches.

This segment also has the distinction of being in remote areas (volcanoes) or sparse population centers (cornfields). This has impacts on data communication systems that we will need to consider later in Chapter 5, Non-IP Based WPAN and Chapter 7, Long-Range Communication Systems and Protocols (WAN).

Agricultural and environmental IoT use cases

Some of the agricultural and environmental IoT use cases are as follows:

  • Smart irrigation and fertilization techniques to improve yield
  • Smart lighting in nesting or poultry farming to improve yield
  • Livestock health and asset tracking
  • Preventative maintenance on remote farming equipment via manufacturer
  • Drone-based land surveys
  • Farm-to-market supply chain efficiencies with asset tracking
  • Robotic farming
  • Volcanic and fault line monitoring for predictive disasters

Energy

The energy segment includes the monitoring of energy production at the source of production to the consumer. A significant amount of research and development has focused on consumer and commercial energy monitors such as smart electric meters that communicate over low-power and long-range protocols to reveal real-time energy usage.

Many energy production facilities are in remote or hostile environments such as desert regions for solar arrays, steep hillsides for wind farms, and hazardous facilities for nuclear reactors. Additionally, data may need real-time or near real-time response for critical responses to energy production control systems (much like manufacturing systems). This can impact how an IoT system is deployed in this category. We will talk about issues of real-time responsiveness later in this book.

Energy IoT use cases

The following are some of the use cases for energy IoT:

  • Oil rig analysis of thousands of sensors and data points for efficiency gains
  • Remote solar panel monitoring and maintenance
  • Hazardous analysis of nuclear facilities
  • Smart electric, gas, and water meters in a citywide deployment to monitor usage and demand
  • Time-of-use tariffs
  • Real-time blade adjustments as a function of weather on remote wind turbines

Smart city

"Smart city" is a phrase used to imply connected and intelligent infrastructure, citizens, and vehicles. Smart cities are one of the fastest growing segments and show substantial cost/benefit ratios especially when we consider tax revenues. Smart cities also touch citizens' lives through safety, security, and ease of use. For example, several cities such as Barcelona have embraced IoT connectivity to monitor trash containers and bins for pickup based on the current capacity, but also the time since the last pickup. This improves the trash collection efficiency allowing the city to use fewer resources and tax revenue in transporting waste, but also eliminates potential smells and odors of rotting organic material.

One of the characteristics of smart city deployment may be the number of sensors used. For example, a smart camera installation on each street corner in New York would require over 3,000 cameras. In other cases, a city such as Barcelona will deploy nearly one million environmental sensors to monitor electric usage, temperature, ambient conditions, air quality, noise levels, and parking spaces. These all have low bandwidth needs versus a streaming video camera, but the aggregate amount of data transmitted will be nearly the same as the surveillance cameras in New York. These characteristics of quantity and bandwidth need to be considered in building the correct IoT architecture.

Smart cities are also impacted by government mandates and regulations (as we will explore later); therefore, there are ties to the government segment.

Smart city IoT use cases

Some of the smart city IoT use cases are as follows:

  • Pollution control and regulatory analysis through environmental sensing
  • Microclimate weather predictions using citywide sensor networks
  • Efficiency gains and improved costs through waste management service on demand
  • Improved traffic flow and fuel economy through smart traffic light control and patterning
  • Energy efficiency of city lighting on demand
  • Smart snow plowing based on real-time road demand, weather conditions, and nearby plows
  • Smart irrigation of parks and public spaces, depending on weather and current usage
  • Smart cameras to watch for crime and real-time automated AMBER Alerts
  • Smart parking lots to automatically find best parking spaces on demand
  • Bridge, street, and infrastructure wear and usage monitors to improve longevity and service

Military and government

City, state, and federal governments, as well as the military, have a keen interest in IoT deployments. Take California's executive order B-30-15 (https://www.gov.ca.gov/news.php?id=18938), which states that by 2030 greenhouse gas emissions affecting global warming will be at levels 40 percent below 1990 levels. To achieve aggressive targets like this, environmental monitors, energy sensing systems, and machine intelligence will need to come into play to alter energy patterns on demand, while still keeping the California economy breathing. Other cases include projects like the Internet Battlefield of Things, with the intent of providing efficiencies for counterattacks on enemies. This segment also ties into the smart city category when we consider the monitoring of government infrastructures like highways and bridges.

The government's role in the IoT also comes into play in the form of standardization, frequency spectrum allocation, and regulations. Take, for example, how the frequency space is divided, secured, and portioned to various providers. We will see throughout this text how certain technologies came to be through federal control.

Government and military IoT use cases

Following are some of the government and military IoT use cases:

  • Terror threat analysis through IoT device pattern analysis and beacons
  • Swarm sensors through drones
  • Sensor bombs deployed on the battlefield to form sensor networks to monitor threats
  • Government asset tracking systems
  • Real-time military personnel tracking and location services
  • Synthetic sensors to monitor hostile environments
  • Water level monitoring to measure dam and flood containment
 

Example use case and deployment

The most effective way to understand an IoT and edge computing system is to start with the use case of a real-world product. Here we will study what the solution is intended to deliver and then focus on the underlying technology. Users and customers will not detail full system requirements, and gaps will need to be derived from the constraints. This example will also illustrate that IoT deployments are a cross-domain collaboration between different engineering disciplines and sciences. Often, there will be digital designers, networking engineers, low-level firmware engineers, industrial designers, human factor engineers, board layout electrical engineers, as well as cloud and SaaS developers. However, the design can't be architected in various silos. Often a design choice in one area could lead to poor performance, bad battery life, exorbitant network charges, or unreliable communication to remote devices.

Case study – Telemedicine palliative care

A provider of in-home care and consultation for elderly and senior citizens intends to modernize their current practice of in-home nursing and nursing assistance care with better, more actionable, and economical solutions to address the growing crisis of costs and patient numbers. Currently the service maintains in-home care with 7-day routine visits to over 500 patients within a 100-mile radius of a metro area in Madison, Wisconsin. Visits consist of everything from delivery of medicine and special care services to measurements of patient vitals. Patients are typically over 70 years old and have no ability to administer any IT infrastructure brought into the home. Additionally, patient homes may not have any Internet connectivity or broadband connection.

Requirements

The provider wishes for a system to provide these minimum feature sets and services:

  • Each patient will be assigned a wearable device to monitor heart rate, blood oxygen, movement, temperature, and steps taken.
  • An additional device(s) will be installed in patient homes to monitor specific patient conditions and vitals such as blood pressure, blood glucose levels, weight, oral temperature, and so on.
  • The system must report data on patient vitals to a central operations dashboard.
  • The system will also remind patients of events such as when to take a certain pill or when to administer a vital test.
  • The system must be able to track status of a patient in the event of power failure.
  • A wearable system is provided with an easily identifiable push button that will signal an emergency situation (such as a fall) to the awaiting operator service. The device will flash to indicate an emergency has been activated. The device will have two-way audio communication with the operator. In the event of a hearing-impaired patient situation, an alternative method will be used to communicate to the patient.
  • The entire network must be able to manage 500 current patients and grow in scale at 10 percent per year.
  • The system must deliver an overall cost savings and ROI of 33 percent within three years of implementation. This key performance indicator (KPI) is measured by reducing in-home nursing and nursing assistance care from three hours per day to two hours per day while increasing the quality of healthcare for patients in the program.

Implementation

Medical IoT and telemedicine are one the most rapidly growing fields of IoT, AI/ML, and sensor systems. With a growth rate year over year (YoY) of 19 percent and a market of $534 billion by 2025, it has garnered much interest. However, we examine this specific case study because of the sizable constraints it places on the system designer. Specifically, in areas of healthcare, the stringent requirements and HIPAA and FDA regulations impose constraints that have to be overcome to build a system that affects patients well-being. For example, HIPAA will require patient data to be secured, so encryption and data security will have to be designed and qualified for the entire system. Additionally, here we examine constraints of the elderly, namely the lack of robust Internet connectivity, while attempting to build an Internet-connected system.

The system will be broken down into three major components:

  • The far edge layer: This will consist of two devices. First will be a wearable device for the patient. The second will be a myriad of different medical-grade measurement tools. The wearable will be a wireless device while the other measurement devices may or may not be wireless. Both will establish secured communication to the PAN-LAN layer component described next.
  • The near-edge PAN-WAN layer: This will be a secured device installed within the locality of the patient home or where they may be cared for. It should be portable but once installed, it shouldn't be used and tampered with by the patient. This will house the PAN-LAN networking infrastructure equipment. It also contains the edge computing systems to manage the devices, control situational awareness, and securely store patient data in the event of a failure.
  • Cloud layer: This will be the aggregation point to store, record and manage patient data. It also presents dashboards and situational awareness rules engines. The clinician will manage a fleet of installed home-care systems through a single dashboard and pane-of-glass. Managing 500 patients with 10 percent growth YoY will present challenges in managing that amount of data quickly especially in emergency situations. Therefore, rules engines will be built to determine when an event or situation exceeds a boundary.

The three layers of the architecture comprise the system from sensor to cloud. The next sections detail aspects of each layer.

The single use case we chose is simply an IoT event from the wearable that must propagate to the cloud for dashboard visibility. The dataflow stretches through all three layers of this IoT use case as shown in the following figure:

Figure 6: Basic dataflow and software components in this use case example. Note the role of the edge computing device providing translation between a Bluetooth device and the cloud over a transport protocol. It also serves the roles of a caching server and encryption agent.

This use case will read from integrated sensors and broadcast the data as BLE advertised packets as a paired BLE device to the edge computer. The edge system manages the relationship to the Bluetooth PAN and will retrieve, encrypt, and store the incoming data in the event of a power or communication failure to the cloud. The edge system has the additional responsibility of converting the Bluetooth data into TCP/IP packets wrapped in an MQTT transport.

It also must set up, manage, and control a cellular communication with a subscription provider. MQTT allows for the reliable and robust transport to the awaiting cloud system (Azure in this example). There data is encrypted over the wire through TLS and then ingested by the Azure IoT Hub. At that point, data will be authenticated and marshalled through the Stream Analytics engine and to the Logic Apps where cloud-based web services will host a dashboard of patient information and events.

Far edge architecture

Let's begin with the far edge and wearable design. For this project, we start with breaking down the user requirement into actionable system requirements:

Use case Choices Detailed description

Wearable patient monitor

  • Wrist strap
  • Neck strap
  • Chest strap
  • Arm strap

Patient-wearable device.

Device should maintain integrity under several environmental parameters such as water submersion, cold, heat, humidity and so on.

Wearable vital monitors

TI AFE4400 heart rate and pulse oximeter

Medical-grade heart rate monitor and blood oxygen sensor

ST Micro MIS2DH Medical Grade MEMS Accelerometer

Movement and pedometer sensor

Maxim MAX30205 Human Body Temperature Sensor

Medical-grade temperature sensor

Emergency call button

A single visible push button with an LED on the unit

The button should depress but not create false positive events. Additionally, a light should flash when an emergency situation is activated. Additionally, a two-way communication can be initiated.

Edge control system

ST Micro STM32WB microcontroller

System to interface sensors and provide PAN communication to PAN-WAN layer. Edge system contains necessary radio hardware and audio codecs.

Microphone

Knowles SPU0410LR5 audio, microphone and amplifier

Two-way communication in event of emergency.

Power system

Encased Li-ion battery in wearable.

The system should have multi-day rechargeable battery life with warnings to the patient and clinician in event of a low-power state. Power system should be rechargeable or replaceable.

Pairing

Bluetooth

Zigbee

Wi-Fi

Need a method to pair and associate wearable attributes within the home PAN-LAN hub.

The intent of the wearable device in a palliative care situation is to be reliable, robust, and durable. We chose medical-grade components and environmentally tested electronics to withstand use cases that may arise in home care. The device also will have no serviceable parts. For example, for this use case, we chose not to burden the patient with recharging the wearable as this procedure may not be followed reliably. Since the project still requires in-home nursing and nursing assistance care, part of the task of the nursing assistance will be to recharge and monitor the health of the wearable.

A system usually starts with the constraints of the embodiment of the components. In this case, a wearable system for in-home health care of the elderly could be in the form of a wristband, neck strap, arm strap and so on. This project has chosen a wrist strap akin to a hospital-style name band that a patient would already have some familiarity with. The wrist strap allows for fitting close to the skin and arteries to allow for collection of health characteristics. Other forms of wearables failed to provide more robust contact. The wrist strap does have significant constraints in size, power, and shape that must contain all the electronics, power supply, and radios described in the following.

From a block diagram perspective, the wearable will consist of as few components as possible to minimize space and weight while conserving as much power as possible. Here we elect to use a very power-efficient microcontroller with a Bluetooth 5 radio (Bluetooth Low EnergyBLE). The BLE radio will serve as the PAN communication to the PAN-WAN hub. BLE 5 has a range of up to 100 meters (or further when LE Long-Range Mode is enabled).

This will be sufficient for in-home care situations where the patient will not necessarily leave.

Figure 7: Wearable computing device for in-home palliative care

Edge layer architecture

The PAN-WAN edge layer is the central edge computer, gateway, and router. In many instances this functionality is performed by a smartphone device; however, in this implementation, we need to construct a system using a different and more economical cellular service plan than what is typically offered for smartphone consumers. Because our scale is 500 users and growing, we decide to build a hub using off-the-shelf hardware components to deliver the best solution for the customer.

The edge computer we chose is an industrial-grade single board computer capable of running an enterprise grade distribution of Linux. The Inforce 6560 acts as a gateway between the Bluetooth 5.0 PAN and a cellular WAN. The system on chip (SOC) conveniently incorporates the following hardware:

  • Snapdragon 660 processor with a Qualcomm Kryo 260 CPU
  • 3 GB of onboard LPDDR4 DRAM
  • 32 GB of eMMC storage
  • One microSD card interface
  • Bluetooth 5.0 radio
  • 802.11n/ac Wi-Fi 2.4GHz and 5GHz radio

Figure 8: Edge system hardware block diagram

The edge computer will also make use of a Bluetooth 5.1 location-tracking angle-of-arrival antenna array. This new standard will allow the edge system to obtain centimeter precision on the location of the wearable and patient within the Bluetooth field. This will allow for tracking the patient's movement, exercise, bathroom functions, and emergency situations.

The edge relies on a failover power system or uninterruptible power supply (UPS). The UPS device will switch from line current to battery in the event of power removal or power outage. It will signal to the edge system that a power event occurred via a USB or serial UART signal. At that point the edge system will communicate back to the cloud management that a power event occurred, and some action may be necessary.

Software architecture

Along with three layers of communication protocols and hardware, there are three different models of software in this relatively simple system. Rather than belabor this use case example with every nuance of the design, including every fault recovery, device provisioning, security, and system state, we will examine the most common usage of delivering patient health data in real time.

The software structure of the wearable device must be compatible with the hardware we have chosen. This implies we chose tools, operating systems, device drivers, and libraries compatible with the architecture and peripherals we are using. We'll start with the wearable, which has the most stringent requirements of code size, battery life, and performance limitations. Since the STM32WB microcontroller is designed as a dual core, we have essentially two systems to manage: the higher-performance ARM M4 core that will run our specific wearable firmware, and a lower-power M0 core that manages the I/O across Bluetooth. We chose a commercial real-time operating system such as ThreadX by Express Logic to allow for a modern development experience rather than a simple control loop which isn't appropriate for this product. We also want to be able to certify the product for medical-grade usage, which is easier when using commercial off-the-shelf operating systems.

The software structure on the wearable is divided into two processes that host numerous threads to manage the wearable display, the speaker and microphone hardware, I/O to the heartbeat and movement sensors, and the Bluetooth stack. The Bluetooth stack communicates across to the M0 core which manages the hardware layers of the Bluetooth radio.

Figure 9: Wearable system software stack divided between two processing cores for application services and IO communication

The edge computer has more processing resources as it must provide for a full TCP/IP stack, PAN and WAN communication and routing, encryption services, storage services, device provisioning, and failsafe firmware upgrades. For the edge system, we chose a Linux Debian-variant system as it provides more features and services than a tightly embedded RTOS. Coordination with the cloud system and all services on the edge computer or on the wearable are coordinated through the "rules engine." The rules engine could be simple "expert system" using custom logic for this customer or use case. A more robust design may use a standardized framework like Drools. Since each patient may need to have a different set of rules, it makes sense to use a dynamic and fungible rule engine that can be uploaded with different patient directives. This is an autonomous top-level supervisor that periodically captures health data, addresses security issues, releases new firmware updates reliably, manages authentication and security, and handles a significant number of error and failure conditions. The rules engine must be autonomous to satisfy the product requirement of the system working without direct control via the cloud.

Figure 10: Edge computer software stack consisting of a number of services managed by a single supervising and autonomous "rules engine"

The cloud layer provides the services of ingestion, long-term data storage, stream analytics, and patient-monitoring dashboards. It provides the interface to the healthcare providers to manage hundreds of edge systems securely through a common interface. It also is the method to quickly provide alerts to health situations, error conditions, and system failures, and provide device upgrades securely. The partitioning of cloud services versus edge services are as follows:

  • Cloud services
    • Data ingestion and management for multiple edge patients and systems
    • Almost unlimited storage capacity
    • A controlled software deployment and updates to edge
  • Edge services
    • Low-latency and real-time reactions to events
    • PAN communication to sensors
    • Minimum connectivity requirements

Commercial cloud services will come with a service agreement and recurring cost, while edge systems for the most part will only incur a single upfront hardware and development cost.

When considering the cloud components, we require a service to ingest data from multiple edge devices securely. The data needs to be stored for analysis and monitoring. The cloud services should also include a method to manage and provision edge installations. Finally, we look for a method to get real-time data from the patients and display it to qualified staff.

For this project, we have chosen to use Microsoft Azure IoT as the cloud provider to manage this large installation and allow for growth and scalability. Azure IoT provides an architecture shown in the following illustration:

Figure 11: Typical Microsoft Azure IoT software stack and cloud architecture

Microsoft Azure IoT software architecture is usually consistent between designs at least on the front end of the IoT Hub. Data will stream from various authenticated sources to the Azure IoT Hub. This is the cloud gateway and capable of scaling to very large IoT installations. Behind the scenes, IoT Hub is a collection of data center processes and services listening and responding to incoming events.

The IoT hub will route qualified streams to the Stream Analytics engine. Here data will quickly be analyzed in real time as fast as data is capable of being ingested. Data can be marshalled to business intelligence services and stored for long persistence in an Azure SQL database and/or moved to a service bus. The service bus responds to events and faults in the form of queues to allow the system to respond to them. The final component in our architecture is the cloud "glue" layers which route data to the IoT device (Logic App Dynamics to Azure) or respond to incoming data (Logic App Azure to Dynamics). Microsoft Dynamics 365 interfaces as a logic app and allows for visibility of IoT events, creation of dashboards, web frameworks, and even mobile and smartphone alerting.

This use case is but a fraction of the actual functionality to make a shipping commercial product. We have left off significant areas such as provisioning, authentication, error conditions, resilient firmware upgrades, system security and root of trust, failover conditions, audio communication, key management, LCD display work, and the dashboard control systems themselves.

Use case retrospective

What we have shown in this very brief introductory use case is that IoT and edge computing requirements for enterprise and commercial designs involve many disciplines, technologies, and considerations. Attempting to trivialize the complexity of bridging Internet connectivity to edge systems with modern expectations of performance, reliability, usability, and security can end in failure.

As we have seen in the abbreviated medical wearable use case, our design involves many interoperable components that comprise a system. It is essential that an architect responsible for an IoT system should be knowledgeable to some level of these system components:

  • Hardware design
  • Power management and battery design
  • Embedded systems design and programming
  • Communication systems, radio signaling, protocol usage, and communication economics
  • Network stacks and protocols
  • Security, provisioning, authentication, and trusted platforms
  • Performance analysis and system partitioning
  • Cloud management, streaming systems, cloud storage systems, and cloud economics
  • Data analytics, data management, and data science
  • Middleware and device management

This book is designed to help the architect navigate through the myriad of details and options for each of these levels.

 

Summary

Welcome to the world of the IoT. As architects in this new field, we have to understand what the customer is building and what the use cases require. IoT systems are not a fire-and-forget type of design. A customer expects several things from jumping on the IoT train.

First, there must be a positive reward. That is dependent on your business and your customer's intent. From my experience, a 5x gain is the target and has worked well for the introduction of new technologies to preexisting industries. Second, IoT design is, by nature, a plurality of devices. The value of IoT is not a single device or a single location broadcasting data to a server. It's a set of things broadcasting information and understanding the value the information in aggregate is trying to tell you. Whatever is designed must scale or be scalable; therefore, that needs attention in upfront design.

We have learned about the segments of IoT and the projected versus actual IoT growth rates. We also have explored a single commercial use case and seen that IoT and edge computing span multiple disciplines, technologies, and functionalities. These mechanics of developing a commercially viable IoT and edge computing system will require the architect to understand these various segments and how they interrelate.

We now start exploring the topology of an IoT and edge computing system as a whole and then break down individual components throughout the rest of the book.

About the Author

  • Perry Lea

    Perry Lea is a 30-year veteran technologist. He spent over 20 years at Hewlett-Packard as chief architect and distinguished technologist of the LaserJet business. He then led a team at Micron focusing on emerging compute using in-memory processing for machine learning and computer vision. At Cradlepoint, he pivoted the company into 5G and the IoT. He then co-founded Rumble, an industry leader in edge/IoT products. He is the principal architect for Microsoft’s Xbox business and works on emerging technologies and hyperscale game streaming, and has authored 40 patents, with 30 pending.

    Browse publications by this author

Recommended For You

Artificial Intelligence with Python - Second Edition

New edition of the bestselling guide to artificial intelligence with Python, updated to Python 3.x, with seven new chapters that cover RNNs, AI and Big Data, fundamental use cases, chatbots, and more.

By Alberto Artasanchez and 1 more
The Complete Edition - Software Engineering for Real-Time Systems

Adopt a diagrammatic approach to creating robust real-time embedded systems

By Jim Cooling