Overall technological advances have contributed to the fact that electronic and other devices become smarter with the ability to produce a large amount of data. In addition, the networking of computers and the Internet has enabled data exchange in both local and Geo-global environments.
From the real-world perspective, including elements of use by ordinary observers such as vehicles, trains, planes, lights, clocks, parking garages … appears a common denominator in the world of smart devices and it’s called the Internet of Things (IoT). It keeps the basic idea of connecting all devices to each other. Internet of things can be understood as the natural evolution of the web because it connects information technology and many other operational technologies. This technology connects more than ordinary devices – it connects all devices that contain such a collection of sensors or any device that generates digital data including humans as an overall data producer. By connecting and networking on a shared thread, the Internet of Things is becoming a based machine that uses data streams as its fuel to build a novel ecosystem based on a few elements.
Things include all sorts of existing devices that are able to produce data directly or by transforming an analog device signal into useful digital data from. Furthermore, within the ecosystem, things are actually virtual copies of objects made up of different forms of data.
Data are digital representation of the various physical forms, objects and even minds thus enabled to be exchangeable, processed by machines and converted into useful information used by human beings as intelligent decisions.
Humans as end node data producers and connected to the Internet become an active actor of the IoT ecosystem by leveraging process of data producing, transforming and sharing.
Processes as a common connectivity and communication descriptor lead us to the overall IoT ecosystem which is able to ingest data autonomously and act as much as self-configured structure.
By representing physical objects and processes by its virtual copies, the IoT ecosystem is limited by a level of fragmentation. Since this is an anomaly, even the ecosystem must be overcome in some way. It is also necessary to allow unhindered data flow and, therefore, the evaluation of the events in the system of independent decision-making. For this, it is necessary to define the structure, i.e. the rules of behavior and functioning of the elements within the IoT ecosystem. Since all of this already coincides and is functionally expressed through already existing technological postulates, it is easiest to define all this through a unique, i.e. the reference architecture of the IoT ecosystem.
The process of forming the reference architecture of the IoT ecosystem is not at all simple. The reason for this is the high dynamism of the elements inside the system, as well as the rapid conglomeration of new solutions in terms of data processing throughout the entire cycle.
In addition, the integration of artificial intelligence into building elements leads to a greater degree of activity and the occurrence of events that can be independently triggered and controlled within the system. This increases the need for efficient scaling of processes and the participation of all other information-science actors within the ecosystem.
In the professional literature and also in practice, there is a more diverse review of the reference architecture of the IoT ecosystem. Starting from those proposed by various equipment manufacturers, IoT elements, platforms of both conventional IoT systems and industrial IoT, there are also references architectures proposed by various consortia and official technical-scientific bodies within governments and intergovernmental organizations.
From this perspective, considering the transformation of existing business processes into the active IoT ecosystem, only some views and perspectives will be discussed without a deeper analysis of all the reference architecture and below are some simplified views of the reference architecture reviews from the software development perspective within the IoT ecosystem.
The Layers of the IoT Architecture
One of the basic and simplified models of the reference architecture is the so-called Conventional IoT architectural model – Three layer IoT Architecture.
Perception layer belongs to the world of sensors, actuators and smart devices. At this level, data production is done. In addition to the active and passive components in this layer, the so-called “edge elements” which, along with “Things” are data producers of the IoT ecosystem. Also, communication from this Layer to the outside world is directed to the cloud environment and is managed using standard network infrastructure as well as localized RF – radio frequency networks.
Network layer covered the area of gateways and routers. Here are most elements of the IoT ecosystem – physical devices or software program between clouds and controllers, sensors and smart devices. Their role is also to discover, i.e. detects intelligent devices in the perception layer and translates them into the network and establish coordination with the applications. If they are hardware devices, they are mostly Plug & Run devices with a simple configuration via a web interface. In some descriptions this layer is also called a Transport layer.
Application layer is a part of the cloud and servers. Functionally, it’s usually some of Cloud services like AVS or Google Cloud, server farms, or even a local remote server company (on premise). The application layer also includes powerful servers and databases that allow large IoT applications to easily integrate, use spanking data storage services with high-speed data processing. The data thus available are easy to filter, build analytics, and API’s user-friendliness. The benefit of this is reflected in an improved and more advanced business logic. In addition, the security of the IoT ecosystem itself with advanced warning systems and real-time monitoring increases.
Support layer was added to the list later as a necessity to give more effective and transparent communication between the application layer and networking layer as contains the services by using APIs.
Five layer IoT Architecture
Another IoT architectural model, is built as an extension of three layer approach. Two more layers are added: a Business and a Processing layer.
With two new layers, IoT architecture still belongs to cloud-centric based architectural model. It turns out that most of the operations or data transfer and processing are performed on cloud or remote servers.
The introduction of a business layer as a top layer is justified by the multifaceted user requirements. The complexity of the business logic in the process of creating the IoT system has sidelined the process of data management in the application layer so that the coordination of business logic is shifted to the new layer. This has led to more stable architecture in terms of the sustainability in perception of connecting different technologies and diverse business areas.
An important element of improving the IoT ecosystem at the architectural level was achieved by introducing the processing layer. As data production increased, IoT systems became cumbersome and difficult to filter data. The layer is often processed, i.e. filtering mass sets of data and data from the Big Data category, in such a way that the processing layer can be found at more than one point within the IoT ecosystem, which leads to resource optimization and more efficient processing and filtering of large amounts of data which is very important for realizing real-time processing as an important factor of efficiency.
Seven layer IoT architecture
Seven layers of IoT architecture is the one most commonly used by users (referred by) when attempting to explain IoT ecosystem appearance and its structure.
The things – in order to realize one IoT environment, i.e. the ecosystem needs to have a variety of devices, sensors and controllers that enable their interconnection. This also means that the end point of an IoT system must have connected devices – besides standard sensors, the actuators also include smartphones, micro-controllers, computers, etc.
Layer two – Connectivity/Edge represents the environment and the place where all connections are made before the exchange of data within the IoT ecosystem. It defines all communication protocols, and establishes a network for Edge computing. Therefore, this type of architecture tells us that this is a distributed architecture and that data are processed on the edge of the network.
The third layer, global infrastructure, is usually a layer that relies on cloud infrastructure. This is because most IoT solutions rely on the integration of cloud services. Observed from the business perspective, this is an inevitable solution in recent times because the cloud provides a complete upgrade to the customer’s perspective.
The fourth layer, the data ingestion, the data entry layer. This is inevitably thought of Big Data, as well as the cleansing and data store. Also, data streaming processes are present in this layer as a building element of data ingestion.
Layer five belongs to the data analysis. This is of course related to the processing of data in order to prepare the report, data mining, the implementation of machine learning, etc.
In the sixth layer, the so-called application layer, user applications is stored depends on the purpose and needs of the user. Different context-elements are also thing-specific and it is therefore necessary to have a defined application layer, otherwise they cannot be standardized within the IoT ecosystem. On the other hand, this means that this layer is where the integration of users and objects from the lowest layer of architecture takes place. In some studies on the IoT architecture, this layer is also called application integration layer where the same layer can be viewed as a service layer with the implementation of the UI at the top.
The seventh layer is represented by people and processes. This includes all business entities as consortium of IoT ecosystems and, at the same time, the actors involved in decision making on the basis of data obtained from the IoT ecosystem, with the help of all the structures that were previously mentioned in architecture.
IoT architecture based on Fog computing paradigm
Increased need for data analysis in real systems and especially in Smart-Grid environments, and also monitoring and pre-processing in the local IoT environment has led to the need to introduce a new layer between the physical layer (where the data producers are located) and the transport or getaway layer. Fog computing has been introduced here as an extension of Cloud and network services to solve this gap on IoT cluster. That’s why it’s logical, that’s similar to Cloud, and the fog is directed at the end user and complements the possibilities of storing data as well as more efficient processing with the use of application services. Furthermore, security and privacy are essential building blocks that greatly contribute in particular to the process of attack analysis and situations where the danger is marked in the system as a man-in-the-middle.
Fog architecture can have more advantages. One of the most important is optimization, i.e. reduced need for information flow in the getaway-cloud relationship. At the same time, it adds a special note in a significant advantage in the real-time performance of the system in general. Another of the interesting possibilities that gives the introduction of a fog layer in the IoT ecosystem is that the fog layer can have direct communication with another fog and thus create a mesh that avoids the use of cloud resources which in some cases of specific IoT solutions can have advantages and contribute to efficiency.
Edge computing IoT architecture
Edge computing is closely associated with the aforementioned Fog computing. The main goal of Edge is that the data processing and functional processing capabilities are redirected to the edge elements of the network environment within the IoT ecosystem.
Inside the Edge environment, all data processing takes place on the physical perception layer itself, or directly on a smart device or on an IoT device collector. All these edges within the Edge can be performed independently on their own layer or in combination with other fog or edge layers from the surroundings of the IoT ecosystem.
Technological component advancements and embedded extensibility contribute to increasing the amount of information directly to the data source, i.e. in elements that produce data. This increases the power of the Edge itself, i.e. the endpoint of the IoT ecosystem, which contributes to conformity and enhances the functionality of other layers such as a transport layer, storage, pre-processing and, of course, monitoring stably reduces the need to transmit a huge amount of data to the cloud. By doing this, the application layer releases successful processing and directs to rapid filtering and analytical processes which leads to the IoT system approaching true processing and predicting the process in real-time.
As an important benefit, Edge allows a greater level of localized latency reduction than any previous one. This is also facilitated by the potential decentralized connection of the IoT ecosystem elements in a global environment. At the same time, the privacy of data is protected by the system and users and globally increases the security of the IoT ecosystem.
Hybrid IoT Architecture
Depending on the project requirement, the architecture of the IoT system can also be constituted as a mixture of Cloud-Fog-Edge. This is particularly advantageous in the context of challenging business goals where it is difficult to meet customer requirements for some of the standard architectures. This combination, i.e. Hybridization is usually called nested.
Edge computing layer: Performs observation and recording user interactions and forwards the feed to the Fog node. From the real-time control signals from the Fog Nodes, the intelligence of the operation is performed directly at the node level.
Fog computing layer: All current data is stored in temporary memory. The control and analytics required to run in real time are based on the application core rules from the cloud.
Cloud computing layer: Performs aggregation of data from all Fog nodes and performs analytical processes on large datasets. In addition, it has the task of forwarding rules for applications execution of Fog Nodes.
The image above shows a nested structure, or a hybridization example of the IoT architecture.
How to follow reference IoT architecture
There are several initiatives and attempts to bring standards regarding the architecture of the IoT ecosystem. The goal of standardization is to achieve the interoperability of the IoT system at the level of platform vendors, research institutions and collaborative initiatives with legislative bodies. This is a process that continues and it is likely that the story will continue to be opened in the future, because the technological advance take place in the high speed and dynamics of the appearance and implementation of increasingly smart devices on the IoT scene.
Some of the reference IoT architectures / standards:
P2413 – IEEE Draft Standard for an Architectural Framework for the Internet of Things (IoT) – IEEE standardization project aims to identify commonalities across IoT domains (including intelligent transport systems, smart cities, manufacturing, smart buildings, smart grid, healthcare…).
Industrial Internet Reference Architecture (IIRA) – specifically developed for industrial IoT applications by the Industrial Internet Consortium (founded in March 2014 -AT&T, Cisco, General Electric, IBM, and Intel).
Internet of Things – Architecture (IoT-A) – IoT reference model and architecture developed through European Union lighthouse project (2013).
Besides general purpose IoT architecture reference models, in the world of various IoT domains and industry verticals exist more specific architecture model signed by different vendors and referenced as platform-centric reference architecture, like:
Designing IoT solution and building ecosystem is very complex and referencing to some specific architecture approach is very uncommon. This depends on the nature of the project itself or the specificity of the design requirement, so that the model can be selected as one of the previously discussed or as a combination of several models. The ultimate goal of referencing architecture is to achieve a greater degree of scalability of IoT ecosystems, end-to-end security by design, automation operations and achieving interoperability both in conceptual and practical terms, as this increases connectivity and communication within the IoT ecosystem thus establishes a general project success.