Technology Internet of Things (IoT) Architecture: Layers Explained

Technology

Internet of Things (IoT) Architecture: Layers Explained

November 13, 2022

The Internet of Things (IoT) will add $5.5 trillion to $12.6 trillion of value to the global economy by 2030. Given that IoT devices include everything from smart light bulbs in homes, to critical sensors at power stations, this figure is not surprising. The challenge, however, is ensuring that such a wide array of devices work together harmoniously. This is where Internet of Things (IoT) architecture comes in – including its layers, systems, and devices.

IoT architecture is the structure enabling internet-connected devices to communicate with other devices. Most IoT architecture models include 3 to 7 sets of functional components, or “layers”, such as perception (e.g., sensors), transport (e.g., Wi-Fi), and application (e.g., software) layers.

Dgtl Infra provides insights into the elements that form IoT architecture, why IoT architecture matters, as well as some of the challenges and future prospects in the field of IoT architecture.

What is IoT Architecture?

IoT architecture refers to the many ways that IoT devices are structured to meet user needs. Based on complexity, IoT system elements are grouped into 3 to 7 layers, each with its own role. Notably, IoT architecture lacks standardized protocols, raising compatibility, security, and other challenges.

According to McKinsey, there will be more than 43 billion IoT devices globally by 2023. These billions of devices are already changing our world – from letting doctors monitor patients remotely, to helping oil companies prevent spills – and they will change our lives even further in the coming years. Underpinning all of this growth is IoT architecture.

Below is an illustration of some of the different layers in IoT architecture:

Layers in Internet of Things IoT Architecture — Click here for a larger version of this image.

What are the Layers of IoT Architecture?

IoT architecture can comprise up to seven layers, which are known as the perception, transport, edge, processing, application, business, and security layers.

1) Perception Layer

The perception layer of an IoT system architecture, also known as the device layer, consists of multiple elements – sensors, cameras, actuators, and similar devices that gather data and perform tasks.

For example, an IoT sensor used on an automotive assembly line can conduct a quality control check on a nearby robot. Each time the robot assembles a fuse box, the IoT sensor checks whether the robot has placed the fuse in the correct position by detecting the color coding of the different fuses.

2) Transport Layer

The transport layer of an IoT system architecture transmits data from multiple devices (e.g., on-site sensors, cameras, actuators) to an on-premise or cloud data center.

As a first step, IoT gateways must convert the incoming input from analog to digital format. Next, the gateway may employ any one of a range of data transfer protocols (DTPs) to transmit the data to an on-premise or cloud data center.

Significant factors that determine the choice of data transfer protocol (DTP) are:

Amount and type of data to be sent
Desired speed and interval of transmission
Reliability of network connection
Power consumption during data transmission
Data and network security
Communication among edge devices

The different DTPs used in IoT networks are characterized by varying advantages and disadvantages with respect to the above factors. Below are some of the most diverse and widely used IoT protocols:

MQTT (Message Queue Telemetry Transport)

MQTT is a lightweight protocol with publish/subscribe interaction schemes, originally designed by IBM. It has come to be the most widely used protocol in the IoT domain due to its open-source nature and suitability for devices located in remote areas with poor internet connectivity.

Modbus

Originally designed for use with the programmable logic controllers (PLCs) of Modicon (now Schneider Electric), the data communications protocol Modbus is a preferred method for connecting a supervisory computer to remote terminal units (RTUs) in IoT systems, following the supervisory control and data acquisition (SCADA) model.

AMQP (Advanced Message Queuing Protocol)

Spearheaded by JPMorgan Chase, one of the largest banks in the U.S., the AMQP protocol was primarily developed for use in transmitting data within the financial services sector. One of the strengths of the AMQP is its in-built security framework that uses components like transport layer security (TLS) and simple authentication and security layer (SASL).

PROFINET (Process Field Network)

Developed and supported by PROFIBUS & PROFINET International (PI), an automation community based in Germany, the Ethernet-compatible PROFINET protocol has been widely adopted in industrial automation systems requiring communication among multiple edge devices, machinery, and software systems.

CAN (Controller Area Network) bus

Originally designed by German engineering and technology company Bosch, CAN bus was designed for use in the automotive industry, where it enabled different devices and sensors within a vehicle to communicate directly, bypassing the requirement of an intermediary computer.

Subsequently, CAN bus has been adapted for a wide variety of two-way device communication uses, including in maritime vessels, construction equipment, lighting control systems, elevator and escalator controls, amongst others.

EtherCAT (Ethernet for Control Automation Technology)

The EtherCAT Ethernet-based protocol was initially developed by German industrial automation company Beckhoff, for systems requiring real-time updating of data. Supported by the EtherCAT Technology Group (ETG), an industrial collective with nearly 7,000 member organizations, EtherCAT is one of the most widely used IoT gateway protocols.

Other Data Transfer Protocols (DTPs)

Several other data transfer protocols (DTPs) exist, such as Constrained Application Protocol (CoAP) and Data Distribution Service (DDS), that are also used extensively in industrial as well as non-industrial IoT applications, including domestic lighting, security, and smart healthcare devices.

3) Edge Layer

As IoT networks grow in scale, latency becomes one of the main performance challenges, as numerous devices connecting to a hub end up congesting the network. By enabling data processing and analysis as close to the source as possible, edge computing addresses these problems – which is handled through the edge layer of an IoT system architecture.

One feature that all IoT edge devices have in common is that they are capable of transmitting what they detect, in the form of data packets, to nodes that then process the data further. Some ‘smart’ edge devices are additionally programmed to halt the target process – or initiate some damage control measure – upon detecting a serious anomaly.

IoT edge devices are typically designed to work smoothly with other devices from different manufacturers, which is important for the proliferation, at-scale, of edge devices in an IoT system.

Data Pre-Processing

Data pre-processing is one of the fastest-growing features of IoT systems and has the potential to detect and resolve problems quickly, while cutting down on the cost of transmitting large volumes of data.

On the other hand, sophisticated processors drive up network layer installation and maintenance costs, and pre-processing architecture carries the additional risk of filtering out valuable IoT device data before it can be processed further.

Pros and Cons – Data Pre-Processing at IoT Gateways

Below is a summary of the pros and cons of enabling data pre-processing at the network level:

Pros	Cons
Immediate insights	High cost of edge processors
Reduced data transmission costs	Remote data security
Local resolution of problems	Potential loss of data insights

4) Processing Layer

A fundamental component of an IoT system architecture is its processing layer, also called the middleware layer, which typically leverages many connected computers simultaneously, in the form of cloud computing, to deliver superior compute, storage, networking, and security performance.

Particularly, the processing layer within an IoT system architecture is responsible for analyzing input data to generate new insights, useful predictions, and timely warnings.

An IoT system typically handles huge volumes of data, generated by numerous edge devices, at multiple sites on the edges of the network. The ‘middleware’ of the processing layer utilizes a three-stage approach to prepare this data for the application layer:

Data Accumulation: middleware correctly identifies and assigns different data types to the appropriate storage. Unstructured data, such as audio and video streams and images, typically require more storage space and are housed in data lakes. Whereas structured data, comprising instrument readings, log values, and measurements (telemetry data) are more space-efficient and are stored in data warehouses
Data Abstraction: involves aggregating data from multiple sources, as well as ensuring that data is converted into a format that can be “read” by the software of the application layer
Data Analysis: employs machine learning (ML) or deep learning algorithms, which are specialized in detecting patterns within large and seemingly random data sets

5) Application Layer

The application layer of an IoT system architecture involves decoding promising patterns in IoT data and compiling them into summaries that are easy for humans to understand, such as graphs and tables. Programs for device control and monitoring, as well as process control software, are typical examples of the application layer of IoT architecture.

6) Business Layer

Patterns decoded at the application level can be used to further distill business insights, project future trends, and drive operational decisions that improve the efficiency, safety, cost-effectiveness, customer experience, and other important aspects of business functionality. Indeed, all of this can be accomplished at the business layer of an IoT system architecture.

Case Study – IoT System Architecture – Business Layer

Celli Group, an Italian producer of drink and beer dispensing equipment, decided to use IoT to address a systemic problem: bar operators were typically unable to assess the health of their dispensing equipment, nor track inventory efficiently, leading to inconsistent product quality as well as missed sales opportunities.

By utilizing Microsoft Azure and PTC’s industrial internet of things (IIoT) platform called ThingWorx, Celli developed IntelliDraught, a retrofit system for turning bar dispensers into ‘smart’ dispensing systems. Specifically, IntelliDraught enables bar operators to collect and send data to a processing system to obtain valuable information on their dispensing equipment’s status, on the quality of the dispensed beverage, and on consumption habits.

In turn, Celli helped bar operators increase customer satisfaction by 27% and sales by 16%. In addition, Celli uses the data generated by IntelliDraught to uncover new insights about beer consumption, bar inventories, and drinking behavior.

7) Security Layer

Security is one of the most important requirements for an IoT system architecture. Ironically, it also happens to be one of the key challenges facing IoT architecture, and IoT devices themselves. Broadly, the IoT security layer comprises three main aspects:

Equipment Security: involves the actual IoT devices, and protecting these endpoints from malware and hijacks
Cloud Security: with most IoT data being processed in the cloud, cloud security is crucial to prevent data leaks
Connection Security: focused on securing data transmitted across networks, primarily with encryption. The transport layer security (TLS) protocol is considered the benchmark for IoT connection security

Challenges of Internet of Things (IoT) Architecture

In addition to security vulnerabilities, the biggest challenges in Internet of Things (IoT) architecture are the obstacles hindering cross-compatibility, connectivity, and mobility, as well as the lack of standardized IoT protocols and languages.

Internet of Things (IoT) Architecture Examples

Examples of Internet of Things (IoT) architecture help explain how these technologies work in different contexts:

1) Internet of Things (IoT) Architecture in Aviation

At any given moment, busy airports have hundreds of vehicles, including aircraft and ground vehicles, moving around. These movements often need to be conducted even during poor weather and low visibility conditions, across areas that span several square miles. In such a context, IoT systems help airports improve operations by reducing disruptions, improving efficiency, and enhancing safety.

Using this example, an airport’s IoT system architecture incorporates the following layers:

Perception Layer: a variety of sensors and devices, such as radio-frequency identification (RFID) tags and GPS sensors, are attached to both vehicles and ground-based locations. These IoT perception layer devices generate data such as vehicle position, movement, distance, speed, traffic frequency, wind direction, visibility, humidity, and more
Transport Layer: this newly generated data is transmitted to one or more servers via an array of technologies in the transport layer, such as 5G, Wi-Fi, and EtherCAT
Processing and Application Layer: hardware and software in the processing and application layer helps convert massive amounts of this raw data into useful, actionable, human-readable data, such as alerting crews and controllers to stranded vehicles or other hazards
Business Layer: all of this actionable data is analyzed to make business decisions that can reduce delays, enhance safety, and minimize emissions, through operational improvements such as better routing and traffic management

2) Internet of Things (IoT) Architecture in Manufacturing

In the early 2010s, power tool manufacturer Stanley Black & Decker employed a simple but ingenious IoT solution that made extensive use of the IoT edge layer. By tagging the products passing through its manufacturing facility with RFID tags, Stanley Black & Decker pinpointed where problems were occurring on their factory floor using a real-time location system (RTLS) tracker, manufactured by AeroScout Industrial, that connected with the factory’s existing Cisco wireless routers.

Overall, these measures increased the efficiency of Stanley Black & Decker’s manufacturing process for power tools from 75% to 95%+, while lowering product defects detected during quality control by 16%.

Amazon Web Services (AWS) – Internet of Things (IoT) Architecture

AWS IoT is the IoT platform of Amazon Web Services (AWS) that simplifies the management and development of IoT solutions at any scale. Specifically, AWS IoT uses the HTTP, MQTT, and MQTT over WebSocket Secure (WSS) communication protocols to securely connect IoT devices to applications and services running on AWS or other cloud platforms.

AWS IoT offers a range of options for device-specific software development kits (SDKs), including in JavaScript, Python, iOS, Android, Arduino, and Embedded C.

Another feature of AWS IoT is the declarative rules engine, which allows IoT traffic to be transformed and directed to a specific location or endpoint. Real-time analytics can also be implemented via applications built within Amazon’s Kinesis Client Library (KCL).

Microsoft Azure – Internet of Things (IoT) Architecture

Azure Internet of Things (IoT) is Microsoft’s managed platform of cloud services that connect, monitor, and control billions of IoT devices. Particularly, Azure IoT supports common communication protocols such as HTTP, MQTT, and AMQP and allows for bi-directional communication between devices and applications.

Microsoft’s Windows for IoT brings power, security, and manageability to the Internet of Things through its operating systems, such as Windows 11 IoT Enterprise, which are exclusively dedicated to running IoT devices. As such, Microsoft provides an end-to-end solution to manage every layer of any IoT system architecture, from device to application.

To this end, Azure IoT also supports Apache Storm, a distributed stream processing computation framework, for real-time analytics. Supported device software development kits (SDKs) include JavaScript, Java, C, Python, and the .NET framework.