Many businesses are jumping on the Internet of Things bandwagon, and turning to IoT consultancy firms. A recent research from Facts & Factors shows that the global IoT market will reach $1,842 Billion by 2028. This growth rate is 24,5%.
It is difficult to roll out IoT. Beecham Research found that 75% of IoT projects fail or are not up to standard. Lack of planning and technical difficulties are common reasons for this.
It is important to have a plan in place for your IoT architecture before you start. This will help avoid the possibility of it going wrong. This blog post will provide insight into the key components of IoT architecture. We also show you how to create an IoT architectural design using an example project from ITRex.
Let’s get started with the basics.
An IoT architecture is an interconnected combination of hardware and software components. This creates a smart cyber-digital system. These components can be interconnected to form a foundation for an IoT solution.
Before we get into the details, let us clarify: There is no single solution for designing IoT architectures. The basic layout remains the same regardless of the solution.
Common data-driven IoT applications rely on a standard IoT architecture spanning four layers:
However, edge processing has been gaining prominence in more connected systems. This means that an additional layer to the traditional four-tier architecture is being added. The percentage of activities performed at an edge depends on the implementation, but it often includes enabling connectivity as well as filtering and aggregating, security, and processing the incoming information.
This is an example of architecture that could be used to create a common IoT solution with edge analytics:
The device layer includes all types of smart, connected devices and non-electronic objects, which are enhanced with sensors, cameras, and actuators.
Sensors receive data from the outside world and convert it to electrical signals that can be used by computers. IoT sensors come in a variety of sizes and purposes. They can record all kinds of information, including temperature, motion, humidity, and many other variables.
The processing center sends commands to actuators, which in turn make the connected devices respond. An actuator receives a command and makes the device behave in a specific way. For example, a smart lighting system could turn on the lights when a motion is detected near it.
The network layer is comprised of different communication technologies that link the device layer with the IoT architecture’s subsequent layers.
Device connectivity can be either enabled directly or via gateways, depending on the IoT solution. This is often used for legacy devices that are unable to connect directly or when there is a protocol mismatch.
These communication technologies are the basis of modern IoT solutions:
We’ve created a table that summarizes the preferred uses of Edge computing layers to help you get an idea of which communication technology is best suited for your solution.
A gateway, local server, or other edge nodes are all part of an edge processing layer. Edge devices are used to store and process data near their source. They can either bulk-upload data to the cloud at predetermined intervals or send only a portion of the records to the cloud. The edge layer could also process data and filter, aggregate, or encrypt it.
Data processing locally saves time and resources, which would otherwise be required to send all the records to the cloud. This results in lower latency and better performance.
An edge layer can be used for IoT applications that require data to be analysed in real-time and need to have built-in security and scalability. This includes smart cars, CCTV systems and medical IoT systems.
Here is where most of the data collected by IoT devices goes. The service and application support layer are used to collect, process, store, and analyze data. Here, two essential processes take place:
IoT systems produce huge amounts of data. However, not all data must be used immediately. An IoT architecture might include a data lake that stores all the information generated and sends only the clean and filtered records down to the data management pipeline.
This stage adds the relevant data from external sources to the information from IoT devices. These could be ERPs, EMRs and other enterprise systems.
The application layer processes the data from IoT devices as well as external sources. Analytic algorithms run the data through the analytics algorithm and present the results to the users.
The business requirements for an IoT system will dictate the type of application that is used. These applications can be web- or mobile-based and present visual insights to end users, control IoT devices using actuators, business intelligence tools or advanced analytics solutions that rely on machine learning or artificial intelligence.
Let’s now see how an IoT architecture might look in practice. We will use ITRex’s portfolio to illustrate the unique aspects of building IoT solutions.
Our client approached us with the idea to build a smart mirror that would allow people to train at home just as efficiently as at a gym. A mirror could be used to “watch” someone working out and provide feedback. It can also help with future training plans. ITRex engineers tackled the problem and created an architecture that encompasses everything, from firmware to mobile apps to hardware.
Our architecture was heavily focused on edge computing. Most of the data from the mirror’s cameras and sensors is processed by the device, with only a small amount of statistical information being sent to the cloud.
Kirill Stathevski, ITRex‘s Chief Technology Officer, explains why edge computing is preferred to traditional cloud-based models. The data collected from the mirror’s cameras, adhesive motion sensors and weights are analyzed right at the source. This helps to save time and reduce operational costs. This is the key to designing IoT architectures that work. You have to make choices, test assumptions and choose what works best for your needs.
This is the high-level architecture of the solution.
A mirror has AI networks that have been trained on extensive video footage. The mirror has built-in cameras that record people working out. After recording, the footage is instantly run through the AI networks, which compare the workout with a reference model. AI engines can generate real-time recommendations about whether or not a person is doing a healthy workout and suggest the necessary improvements, such as weight, technique, intensity, etc.
A trainee uses the mirror to draw on video footage in order to personalize AI networks locally. The quality of the suggestions increases with time.
Kirill says personalization is another reason why we chose an edge-oriented architecture. The local training of the algorithms based on the videos taken in the actual context where the mirror is used yields better results than the cloud-based training that relies on generic content. Privacy is another reason to choose an edge-centered architecture. Data is processed close to the source and is not transferred across the network to be analysed.
Although the architecture is edge-oriented, it also includes the cloud component. Its main purpose is to collect statistical data about the performance and usage of mirrors.
A social mobile app is another component of the solution that allows end users to track their performance, share it, and train together.
It is important to plan for the future if you are serious about adopting IoT. Systems that are poorly designed and not flexible can’t handle complexity and aren’t scalable. However, a well-designed IoT architecture will allow you to plan for the future.