A Simplified IoT Architecture & The Core IoT Functional Stack

The framework is presented as two parallel stacks: the IoT Data Management and Compute Stack and the Core IoT Functional Stack. The goal is not to promote or endorse any one specific IoT architectural framework, but rather to simplify the IoT architecture into its most basic building blocks and use it as a foundation to understand key design and deployment principles applied to industry-specific use cases.

The Core IoT Functional Stack is presented in three layers, allowing for better visibility into the functions of each layer. 

• The network communications layer of the IoT stack involves a significant amount of detail and incorporates a vast array of technologies. The last-mile technologies used in IoT are chosen to meet the specific requirements of the endpoints and are unlikely to ever be seen in the IT domain. The network between the gateway and the data center is composed mostly of traditional technologies that experienced IT professionals would quickly recognize.
  
• The applications and analytics layer of IoT doesn't necessarily exist only in the data center or in the cloud. Due to the unique challenges and requirements of IoT, it is often necessary to deploy applications and data management throughout the architecture in a tiered approach, allowing data collection, analytics, and intelligent controls at multiple points in the IoT system. Data management is aligned with each of the three layers of the Core IoT Functional Stack.

The applications layer of IoT networks is different from the application layer of a typical enterprise network, as it often involves a strong big data analytics component. Security is central to the entire architecture, both from network connectivity and data management perspectives.

The Core IoT Functional Stack

IoT networks are constructed based on the idea of "things," which are intelligent objects that carry out tasks and provide novel interconnected services. These objects are considered "smart" because they utilize both contextual information and pre-set objectives to execute actions. These actions can be self-contained, meaning that the smart object can operate independently without relying on external systems. However, in most cases, the smart object interacts with an external system to transmit the information it collects, exchange data with other objects, or interact with a management platform. The management platform can be utilized to analyze the data gathered from the smart object and direct the actions of the smart object.

From an architecture perspective, multiple components must collaborate in order for an IoT network to be functional.

Things Layer

At this level, the physical devices must conform to the limitations of the environment in which they are deployed, while also being capable of delivering the required information. There are numerous ways to classify smart objects. One architectural classification could be: 

• Battery-powered or power-connected, 
• Mobile or static, 
• Low or high reporting frequency, 
• Simple or rich data, 
• Report range, 
• Object density per cell.

From a network architectural standpoint, your initial task is to determine which technology should be used to allow smart objects to communicate. This determination depends on the way the things are classified. However, some industries may include objects in various categories, matching different needs.

 

Communications Network Layer

In cases where smart devices lack self-sufficiency, they must establish communication with an external system. Typically, this communication employs wireless technology. There are four sublayers in this layer.

1. Access network sublayer:

The access network is the final segment of the IoT network. This is commonly composed of wireless technologies such as 802.11ah, 802.15.4g, and LoRa. The sensors linked to the access network may also be connected via physical cables.

2. The Gateways and Backhaul network sublayer: 

It refers to the specific layer inside a network that handles the transmission of data between other networks and the main network infrastructure. A typical communication system arranges numerous intelligent things inside a specific region, all connected to a central gateway.  

 

The gateway establishes direct communication with the smart items. The function of the gateway is to transmit the gathered data over a long-distance communication system (referred to as the backhaul) to a central station known as the headend, where the data is then analyzed and processed. 

 

The process of exchanging information in this context is a function that operates at Layer 7, also known as the application layer. This is why this particular object is referred to as a gateway. On IP networks, this gateway functions as a router by forwarding data across different IP networks.

3. The network transport sublayer  

is responsible for ensuring successful communication by implementing network and transport layer protocols, such as IP and UDP. These protocols facilitate the connection of various devices and the use of different media.

4. The IoT network management sublayer: 

It requires the implementation of additional protocols to enable the transmission of data between the headend apps and the sensors. Some examples of communication protocols are CoAP and MQTT.

Application and Analytics Layer

At the higher layer, an application is responsible for processing the gathered data. Its role is not only to control the smart objects as needed, but also to make intelligent decisions based on the collected information. These decisions are then used to instruct the "things" or other systems to adapt to the analyzed conditions and modify their behaviors or parameters. The subsequent sections analyze these components and assist you in designing an IoT communication network.

Analytics Versus Control Applications

Multiple applications can help increase the efficiency of an IoT network. Each application collects data and provides a range of functions based on analyzing the collected data. From an architectural standpoint, one basic classification can be as follows: 

 

Analytics application: This type of application collects data from multiple smart objects, processes the collected data, and displays information resulting from the data that was processed. The display can be about any aspect of the IoT network, from historical reports, statistics, or trends to individual system states. The important aspect is that the application processes the data to convey a view of the network that cannot be obtained from solely looking at the information displayed by a single smart object.

Control application: This type of application controls the behavior of the smart object or the behavior of an object related to the smart object. For example, a pressure sensor may be connected to a pump. A control application increases the pump speed when the connected sensor detects a drop in pressure. Control applications are very useful for controlling complex aspects of an IoT network with a logic that cannot be programmed inside a single IoT object, either because the configured changes are too complex to fit into the local system or because the configured changes rely on parameters that include elements outside the IoT object. 

 

An example of control system architecture is SCADA. SCADA was developed as a universal method to access remote systems and send instructions. One example where SCADA is widely used is in the control and monitoring of remote terminal units (RTUs) on the electrical distribution grid.

Data Versus Network Analytics

Analytics is a general term that describes processing information to make sense of collected data. In the
world of IoT, a possible classification of the analytics function is as follows:

 

Data analytics: This type of analytics processes the data collected by smart objects and combines it to provide an intelligent view related to the IoT system. At a very basic level, a dashboard can display an alarm when a weight sensor detects that a shelf is empty in a store. In a more complex case, temperature, pressure, wind, humidity, and light levels collected from thousands of sensors may be combined and then processed to determine the likelihood of a storm and its possible path. In this case, data processing can be very complex and may combine multiple changing values over complex algorithms. Data analytics can also monitor the IoT system itself. For example, a machine or robot in a factory can report data about its own movements. This data can be used by an analytics application to report degradation in the movement speeds, which may be indicative of a need to service the robot before a part breaks.

 

Network analytics: Most IoT systems are built around smart objects connected to the network. A loss or degradation in connectivity is likely to affect the efficiency of the system. Such a loss can have dramatic effects. For example, open mines use wireless networks to automatically pilot dump trucks. A lasting loss of connectivity may result in an accident or degradation of operations efficiency (automated dump trucks typically stop upon connectivity loss). On a more minor scale, loss of connectivity means that data stops being fed to your data analytics platform, and the system stops making intelligent analyses of the IoT system. A similar consequence is that the control module cannot modify local object behaviors anymore.

Data Analytics Versus Business Benefits

Data analytics is undoubtedly a field where the value of IoT is booming. Almost any object can be connected, and multiple types of sensors can be installed on a given object. Collecting and interpreting the data generated by these devices is where the value of IoT is realized.

From an architectural standpoint, you can define static IoT networks where a clear list of elements to monitor and analytics to perform are determined. Such static systems are common in industrial environments where the IoT charter is about providing a clear view of the state of the operation. However, a smarter architectural choice may be to allow for an open system where the network is engineered to be flexible enough that other sensors may be added in the future, and where both upstream and downstream operations are allowed. This flexibility allows for additional processing of the existing sensors and also deeper and more efficient interaction with the connected objects. This enhanced data processing can result in new added value for businesses that are not envisioned at the time when the system is initially deployed.
 

An example of a flexible analytics and control application is Cisco Jasper, which provides a turnkey cloud-based platform for IoT management and monetization. Consider the case of vending machines deployed throughout a city. At a basic level, these machines can be connected, and sensors can be deployed to report when a machine is in an error state. A repair person can be sent to address the issue when such a state is identified. This type of alert is a time saver and avoids the need for the repair team to tour all the machines in turn when only one may be malfunctioning.

---

Happy Exploring!