1.1 Introduction

The Internet of Things (IoT) has connected billions of sensors and devices and created many opportunities that led many businesses to initiate IoT marketing and development budgets. Affordable, smaller and more powerful hardware, ubiquitous connectivity, availability of big data tools and cloud-based services, and huge market awareness are some of the reasons why more and more IoT opportunities and use cases are occurring [1]. To bring these many use cases and opportunities to life and handle the huge amount of data generated by massive number of IoT devices and sensors, a solid infrastructure and architectural model is needed. From a network architectural standpoint, the simplified core IoT functional stack has four main layers as shown in Figure 1.1.

Note that security is an essential element of each building block of this IoT technology stack.

A.  Hardware

Hardware is where the data is collected and includes the smart objects (Things) with built-in sensors to measure physical data, actuators to perform tasks, low cost microprocessor, communication device to receive instructions, send or route data, and a power source (battery, mains, solar, etc.).

The characteristics and type of the smart objects in the hardware layer and their requirements can determine the technologies and protocols of the upper layers in the IoT functional stack. Battery-powered smart objects can have high mobility and require wireless technologies for their communications. However, their transmission range or reporting frequency could be limited. On the other hand, power-connected things are typically static and they are allowed to use technologies and protocols with higher power consumption that can provide longer transmission range and report their data more frequently. The amount and type of data collected by the smart objects and exchanged during the report cycles have an impact of the power consumption and consequently the choice of the connectivity technologies and protocols. For example, the electroencephalogram (EEG) signals collected by a smart helmet are richer compared to the data collected by a simple temperature sensor and hence, require more power and throughput to transmit the data.

Another important characteristic of a smart object is its report range or basically how far the data needs to be transmitted. For example, a vibration sensor installed on the train rails in a rural area may need to communicate with a cellular tower a few hundred meters away. Thus, for such a use case with a long report range, a low power communication technology is required to cover a wide area network.

The aforementioned hardware examples show that in order to design an IoT solution, a first step could be to examine the requirements of the “Thing” in terms of power consumption, report range, mobility, frequency of report and data transmission rate [2].

B.  Connectivity

The connectivity or communications layer connects the smart object (hardware) to the network through an IoT access technology (e.g., WiFi, Bluetooth, etc.). Basically, the connectivity layer is responsible to transport the collected data from the sensors to the Internet through different access technologies with fundamentally different architectures, data transmission rates, report range, power consumption and security requirements. The majority of the existing IoT devices use one of the following forms of data transfer infrastructure shown in Figure1.2 to communicate with the software backend (cloud services) where the data and all connected devices are managed.

Ad-hoc infrastructure

In this unlicensed short-range communication model, the “Thing” is connected to the Internet through a user-provided communications gateway or a vendor managed mesh network. In the user-provided architecture, the connectivity of smart object to the software backend (cloud or data storage infrastructure) is provided by technologies such as Bluetooth or WiFi that operate in the unlicensed spectrum. Although such technologies have high data transmission rates, security, provisioning and device management problems (e.g., different WiFi standards and security models, passwords change etc.) are major barriers to extend these technologies into industrial and enterprise use cases. Moreover, the high power consumption (limited battery life cycle of the IoT devices and higher maintenance costs) and low reporting range of such ad-hoc networks could be other barriers to adopt these unlicensed spectrum technologies in a broader range of IoT use cases.

In vendor managed mesh networks, communication protocols such as ZigBee or Z-Wave are deployed to transmit the collected data into a gateway which then forwards the received data into the software backend usually via cellular networks (4G/LTE). In these conventional costly and power hungry solutions, different libraries need to be set up for each different type of simple or complex smart object to integrate them into the IoT platform [1].

Unlicensed spectrum technologies

LoRa and Sigfox are two low power, widely used and relatively cheap IoT technologies intended for battery operated smart objects with high report range (wide area networks). These unlicensed spectrum technologies target key IoT requirements such as secure bi-directional communications, mobility and localization services; however, their low data rate limits the capability to update the firmware and send complex commands. The maximum data rates for LoRa and Sigfox are 27kbps and 100bps, respectively.   Interference in high volume industrial applications is another issue in deploying these technologies. Moreover, the costly deployment of the technology and the need to set up a physical network infrastructure is another barrier in adoption of these alternative solutions.

Cellular Low-Power Wide-Area (LPWA)

One main important advantage of the cellular technologies over other unlicensed LPWA access technologies is their ubiquitous connectivity. The cellular networks are almost always available even in remote locations. Cellular IoT (CIoT) technologies are highly reliable, secure, and scalable which makes them suitable candidate technologies for massive IoT where a huge number of smart objects (up to 1 million devices per km²) are connected to the Internet. Moreover, cellular communications’ maturity provides global ecosystem interoperability for CIoT technologies.

The 3rd Generation Partnership Project (3GPP) – a global technical body which develops technical specifications for mobile communication system – introduced a suite of two complementary CIoT technologies called enhanced Machine-Type Communications (eMTC or LTE-M), and Narrow band Internet of Things (NB-IoT) in its Release 13 [3].

LTE-M is a low‑power wide‑area (LPWA) air interface in the licensed spectrum that lets IoT devices connect with a data rate up to 1Mbps [4]. Thanks to two innovations of LTE Cat-M1 introduced by 3GPP Release 13, Power Saving Mode (PSM) and extended Discontinuous Reception (eDRx) longer battery lifecycles (up to a few years) are enabled and greater in‑building range is supported as compared to standard cellular technologies such as 3G or LTE Cat 1. The power efficient LTE-M also supports voice functionality via VoLTE (used for applications requiring a level of human interaction), as well as full mobility and in‑vehicle hand‑over (used for asset tracking applications).

NB-IoT uses a small bandwidth for data transmission and reduces the bandwidth to 200 kHz (180 kHz plus guard-band) as compared to 1.4 MHz used in LTE-M [3]. Also, compared to LTE-M, NB-IoT consumes less power (longer battery life cycle) and provides even deeper coverage. The reduced complexity of NB-IoT makes it suitable for low data rate (10’s of kbps), delay-tolerant use cases including sensors and meters.

C.  Software backend

Software backend consists of cloud services that manage the network and the IoT devices. The integration of data and the interface to the 3rd party systems such as Enterprise Resource Planning (ERP) systems are provided by these cloud services. IoT platforms such as Microsoft Azure IoT Hub, Amazon Web Services (AWS) IoT, IBM Watson IoT are part of the central software backend. Event processing and action management, advanced analytics, privacy management, storage/database, device management, and integration with external interfaces are some of the functionalities of the IoT platforms.

D.  Application

The application layer visualizes the collected data from the sensors in real-time and integrates the business systems. This is how the raw collected data from the sensors is turned into value for businesses and companies. Moreover, the behavior of the smart things is controlled by the commands sent from the application layer to the smart objects.

Typically, there are two types of analytics in the application layer that generate two different intelligent views about the IoT system and the IoT network. First, data analytics which provides insight about the IoT system through analysing and processing the data collected by the sensors. Second, network analytics that evaluates the connectivity of the network to maintain the quality of service and prevent future network failures. The IoT network architect determines the depth of the network analytics depending on the use case. For use cases such as environmental monitoring where the sensors occasionally transmit their collected data without the need for an immediate action, a basic network analytics could be sufficient. On the other hand, in mission critical use cases such as connected vehicle, it is extremely impotent to have a detailed view of the network connectivity and performance.