As shown in Figure 1, Ericsson predicts there will be around 28 billion connected devices by 2021, of which more than 15 billion will be connected M2M and consumer-electronics devices . A large 25% share of these will be applications served by short-range radio technologies such as Wi-Fi and Bluetooth, while a significant 20% proportion will be enabled by wide area networks (WANs) that are primarily facilitated by cellular networks.
The IoT revolution offers huge potential value in terms of improved efficiency, sustainability and safety for industry and society. Analysts predict that the total added value of the IoT will be USD 1.9 trillion by 2020 .
There are a wide range of existing IoT use cases, as the market is now expanding toward both Massive IoT deployment as well as more advanced solutions that may be categorized as Critical IoT .
Massive IoT encapsulates all use cases that require many devices at low cost, low energy, and small data volumes applied in slow processes such as a smart buildings and smart agriculture. At the other end of the scale, Critical IoT covers all use cases that operate at high speed and deal with the vital maintenance human lives and industry, such as e.g. health tracking, traffic safety and control, as well as smart grid automation. These devices demand high reliability and availability as well as low latency. Thereby, the data volumes are much higher, but the business value is significantly higher. Specifically, these use cases are enabled by LTE or 5G capabilities. An overview can be seen in Figure 2.
There are, however, many other use cases between the two extremes, which today rely on 2G, 3G, or 4G connectivity.
The IoT market segment comprises several applications widely used in industries and for personal use, as shown in Figure 3.
The potential applications for IoT run into the thousands, with a huge variety of requirements regarding cost, battery life, coverage, connectivity, performance (throughput and capacity), security and reliability.
Some devices will only send a few messages per day – such as status indicators for temperature or tank level – while others may need to transmit a video performance stream to guide a remote repair technician, for example. The difference in throughput requirements is huge. If operators or service providers handle several applications, it may be of great benefit to be able to harmonize communication modules, so that they all use the same underlying radio solution to reduce operational and fault management effort and complexity .
Many higher-value applications will require two-way communications – in other words, an uplink as well as a downlink – to enable monitoring and control of devices in systems like heating, ventilation and cooling plants. The long lifetime of many IoT applications makes it invaluable to be able to perform over-the-air device updates for new functionality or parameter settings.The amount of data sent for such updates can often be more demanding for the network than the monitoring or control application itself.
Relatively simple uplink-focused applications can benefit greatly from a bi-directional link to provide robustness. For example, a connected smoke detector must deliver a smoke alarm with absolute certainty. The ability of a network to provide acknowledgements of a received message enables better fault management and the required level of reliability. Positioning can be used to locate the sensor at the point it failed and simplify operations. For tracking applications, location information is essential.
In applications like personal health monitoring, sensitive information could be reported over the air, which will require strict security. Furthermore, in the case of a medical emergency, it is crucial that the alarm information reaches the control center in time – making QoS and two-way communication vital.
Connectivity is the foundation for IoT, and the type of access required will depend on the nature of the application. Many IoT devices will be served by radio technologies that operate on unlicensed spectrum and that are designed for short-range connectivity with limited Quality of Service (QoS) and security requirements typically applicable for a home or indoor environment. Currently, there are two alternative connectivity tracks for the many IoT applications that depend on wide-area coverage:
Cellular technologies: 3GPP technologies like GSM, WCDMA,LTE and 5G. These WANs operate on licensed spectrum and historically have primarily targeted high-quality mobile voice and data services. Now, however, they are being rapidly evolved with new functionality and the new radio access technology narrowband IoT (NB-IoT) specifically tailored to form an attractive solution for emerging low power wide area (LPWA) applications.
Unlicensed LPWA: new proprietary radio technologies, provided by, for example, SIGFOX and LoRa, have been developed and designed solely for machine-type communication (MTC) applications addressing the ultra-low-end sensor segment, with very limited demands on throughput, reliability or QoS.
One way to segment IoT applications is to categorize them according to coverage needs and performance requirements (such as data speed or latency demands).
The coverage needs of a particular use case may be highly localized (such as a stationary installation within a building), while other use cases require global service coverage (such as container tracking). 3GPP technologies already dominate use cases with large geographic coverage needs and medium- to high-performance requirements.
With new feature sets specifically tailored for LPWA IoT applications, 3GPP technologies are taking a large leap forward to cover segments with low-cost, low-performance requirements too.
Each of the technologies available for IoT connectivity has its own advantages and disadvantages. However, the range of IoT connectivity requirements – both technical and commercial – means cellular technologies can provide clear benefits across a wide variety of applications. These benefits include:
In terms of global reach, cellular networks already cover 90% of the world’s population. WCDMA and LTE are catching up, but GSM will offer superior coverage in many markets for years to come. Cellular networks have been developed and deployed over three decades, and they will be around for the foreseeable future.
The cellular mobile industry represents a huge and mature ecosystem, incorporating chipset, device and network equipment vendors, operators, application providers and many others. The global cellular ecosystem is governed by the 3GPP standardization forum, which guarantees broad industry support for future development.
comes to scalability, cellular networks are built to handle massive volumes of mobile broadband traffic; the traffic from most IoT applications will be relatively small and easily absorbed. Operators are able to offer connectivity for IoT applications from the start-up phase and grow this business with lowTCO and only limited additional investment and effort. Operation in licensed spectrum also provides predictable and controlled interference, which enables efficient use of the spectrum to support massive volumes of devices.
Cellular connectivity offers the diversity to serve a wide range of applications with varying requirements within one network. While competing unlicensed LPWA technologies are designed solely for very low-end MTC applications, cellular networks can address everything from Massive to Critical IoT use cases.
Quality of Service (QoS) mechanisms will be essential for many IoT applications.Cellular systems have mature QoS functionality, and this enables critical MTC applications to be handled together with traffic from sensors, voice and mobile-broadband traffic on the same carrier. QoS, along with licensed spectrum as described above, provides a foundation for long-term Service Level Agreements with a specific grade of service.
Traditionally, the security mechanisms of cellular networks have been based on a physical SIM attached to the device, referred to as a Universal Integrated Circuit Card (UICC). This has also enabled roaming between operators, which has been one of the main factors behind the huge success of mobile networks. The SIM will also be essential in future IoT applications, with SIM functionality embedded in the chipset (eUICC) or handled as a soft-SIM solution running in a trusted run-time environment of the module.
With a straightforward rollout of new software, cellular networks will be able to support the full breadth of applications, ranging from low-end use cases in the LPWA segment, to the high-end segments of in-car entertainment and video surveillance.One network connecting the whole diversifying IoT market will guarantee the lowest possible TCO as well as rapid time to market.
The NB-IoT global technology solution has been standardized according to 3GPPRelease 13 . NB-IoT is a self-contained carrier that can be deployed with a system bandwidth of only 200kHz, and is specifically tailored for ultra-low-power IoT applications. It is enabled using new network software on an existing LTE network, which will result in rapid time to market .
NB-IoT provides lean setup procedures, and a capacity evaluation indicates that each200kHz NB-IoT carrier can support more than 200’000 subscribers. The solution can easily be scaled up by adding multiple NB-IoT carriers when needed. NB-IoT also comes with an extended coverage of up to 20dB, and battery saving features, Power Saving Mode and eDRX for more than 10 years of battery life.
LTE basestations basically support the extensions on the hardware side, LPWAN is almost already built into the basic technology. In many cases, updating the software is sufficient to make the radio systems fit for NB-IoT. Mobile network operators can therefore use existing LTE infrastructure facilities to provide IoT services at a reasonable cost. It is therefore to be expected that NB-IoTwill soon be widely available. Deutsche Telekom, one of the main players in the local NB-IoT market, began upgrading its base stations in the second quarter of2017. The rollout is to be completed by the end of 2018. Vodafone also relies heavily on narrowband technology.
Since the transmission requires only a small bandwidth, several 10,000 IoT endpoints can be supplied with one radio cell. LTE Cat-NB1 transmits in licensed frequency ranges and is characterized by high interference immunity and relatively low latency times compared to other LPWAN technologies. However, these can be 10seconds or more. Another advantage: NB-IoT allows worldwide roaming and thus global scaling of IoT applications.
NB-IoT is designed to be tightly integrated and interwork with LTE, which provides great deployment flexibility. The NB-IoT carrier can be deployed in the LTE guard band, embedded within a normal LTE carrier, or as a standalone carrier in, for example, GSM bands.
A base station dynamically divides the PRBs among the terminals, the so-called User Equipment(UE). This is done in time slots of one millisecond each. Distribution is variable, depending on data volume, channel conditions and quality of service requirements. 3GPP has specified that each LTE-UE must be able to process between 1 and 100 PRBs per millisecond.
For NB-IoT,3GPP has extended the LTE standard and defined a new air interface with Release13: LTE Cat-NB1 UEs transmit and receive on a radio channel that is only 180kHz wide. This width corresponds exactly to that of an LTE-PRB. This limits the achievable data rate per time slot. Depending on the source, the specifications for the maximum downlink and uplink data rate vary between 150 and 250 kbps.OFDMA and SC-FDMA are again used as access methods. This enables lossless embedding of the NB-IoT carrier signal in the broadband LTE signal. A common term for this is in-band operation.
NB-IoT reduces device complexity below that of LTE-M with the potential to rival module costs of unlicensed LPWA technologies, and it will be ideal for addressing ultra-low-power applications in markets with a mature LTE installed base.
Cloud computing until today has mainly laid its focus on storing data generated by IoT devices and using compute cycles to extract value from analyzing this data. In the general public, there is a misconception of what the cloud is and what it means to store and compute in the cloud.
Bluntly spoken, the cloud is an agglomeration of vast data centers, which house a network of powerful connected computers, known as a cluster. The owners of these data centers, such as Microsoft, Amazon, or Google allocate some computing units to be ‘rented’ by individuals or enterprises. Cloud services can be imagined as three layers, each of which are tightly interconnected, as demonstrated in Figure 8.
Infrastructure as a Service (IaaS): This model provides basic IT resources such as computing power, storage or network capacity. The user has control over operating systems and applications, he/she usually must assemble the infrastructure him-/herself from the required computing instances and storage.
Platform as a Service (PaaS): PaaS is a type of service that provides a programming model and developer tools to build and run cloud-based applications. A PaaS provider should automatically provide all the required resources such as computing power, memory, network, middleware such as message queuing or load balancing and databases when deploying the application and scale them according to the requirements (“fabric”). Built-in monitoring functions are also expected, with which the runtime behavior of the applications can be monitored.
Software as a Service (SaaS): SaaS represents the top layer in the cloud model, where the provider makes its own applications available to users. This is where SaaS differs from its predecessor ASP (Application ServiceProvider), where service providers offered applications from other manufacturers for rent. These were usually not multi-client capable, while support for multiple clients is the rule with SaaS.
Finally, the customer of a B2C or B2B SaaS provider does not need to worry about the technical infrastructure or the installation and updates of the application. This is done centrally by the provider so that users always have access to the latest version of the software.
The main advantage of cloud services is that they offer standardized services faster and at a lower price than the companies themselves can with their internal IT. Providers achieve this through a high degree of automation in their data centers and optimal utilization of their resources by a heterogeneous and globally distributed user base. In addition, there is pricing that invites customers to take advantage of services outside peak periods. Development environments, SDKs and middleware services are bug-free and instantly usable for developing and testing new solutions or integrating existing solutions. Users do not have to make the usual initial investments; only running costs are incurred from subscribing to cloud services. In addition, the time required to make IT resources available is considerably shorter compared to traditional procurement, so that companies can react more quickly to new requirements .
Microsoft Azure’s IoT Hub and Amazon’s AWS IoT platform are by nature PaaS that form a gateway for bi-directional communication between IoT terminals on the one hand and cloud services on the other.
At Microsoft for example, communication takes place via so-called endpoints of IP-enabled end devices, and cloud security is now inventing itself optionally via automatic and/or manual routing. IoT Hubsupports multiple messaging patterns such as device-to-cloud telemetry, file uploads from devices, and request-response methods for controlling devices via the cloud. IoT Hub Monitoring helps ensure the integrity of the solution by tracking events such as device creation, device failures and device connections.
In contrast, AWS uses a rule processing engine to handle this bidirectional communication. At the heart of Amazon's IoT platform is the Device Gateway, a communication interface between the terminals and the cloud that transmits relevant messages back and forth between IoT terminals and the control processing engine. Connected devices report their status to a message broker who stores this data in a device shadow and forwards it to the recipients who have subscribed to the associated thread. Applications can then request the required changes to the state of a device .
Amazon’s strengths lie in the massive scalability of its infrastructure and orchestration. AWS scales to billions of IoT devices and trillions of messages. Azure scores with support for predictive maintenance, machine learning (Cortana Intelligence Service and CognitiveServices have no analog in AWS) and visualization using Power BI. One thing is for sure: both providers are equally overpowering each other in inventing new IoT services.
A typical architecture for the integration of IoT devices into the Azure infrastructure is presented in Figure 9.
The architecture consists of three main components:
IP capable devices equipped with an IoT client and the capacity to send messages over pre-required protocols (HTTPS, AMQP, MQTT) may circumvent passing messages to the cloud gateway through an external IoT gateway. This situation arises for many existing IoT devices and low-power devices, which send message via alternative protocols. In general, security against over-the-air MITM attacks can be prevented by token-basedSAS authentication, individual X.509 certificates, or X.509 CA authentication for the exchanged messages.
After messages from specific devices are received in the cloud gateway, they are buffered for storage and routed to the stream processors, which delegate the data to the app backend, to the Analytics & Machine Learning module, and database storing the device states and identities. In a bidirectional message-sending scenario, acknowledgement of device message reception and control commands are executed by sending REST requests via the Provisioning API.
The third category comprises the frontend and manages the interactions and interfaces of a user on a personal mobile device as well as on a business (ERP) system.
With all its mature infrastructure, IoT systems that purely rely on an external cloud providers still carry some disadvantages with it, such as transfer latency after data processing and a certain security risk. New developments in fog and edge computing have been directly tackling these issues.
In an abstract sense, edge computing defines the addition of an additional computing layer between the IoT devices and the cloud server. This (1) minimizes the data transfer latency created by physical distance between cloud server location and device, (2) relieves the cloud server from computation that is only specific to a small number of devices in the network, and lastly (3) adds a layer of security as there is no single cloud server that can be attacked. Conceptually, the computing power is pushed to the edge.
If the data is analyzed directly at the sensors and actuators, i.e. very early in their lifecycle, valuable information can be obtained more quickly. As the processing moves closer to the sensor, innovations in the measurement system software must also ensure that the analysis is moved further in the direction of edge. Future software for edge-based systems will make it possible to quickly configure and manage thousands of networked instruments and transmit a vast amount of analysis and signal processing data to these nodes. Companies must move to smarter, software-based measurement nodes to compete with the amount of analog data they will generate .
Once intelligent systems have captured the data, the next step is to make the data available to the enterprise for efficient management, consolidation and analysis. An enterprise-wide data management and analysis solution that can handle data from multiple sources is essential to deliver the right data to the right people at the right time so they can make informed decisions based on the data.
All devices that can be connected will be connected.
The uptake of massive and critical IoT currently lays the foundation for the technology to skyrocket, and operators have a unique opportunity to drive the implementation of useful IoT applications by offering affordable connectivity on a global scale.
For IoT applications, existing cellular networks offer distinct advantages over alternative WAN technologies, such as unlicensed LPWA. The global reach, QoS, ecosystem, TCO, scalability, diversity and security of cellular networks are all vital factors that can support the fast uptake and success of IoT. Enabled through new software in existing legacy networks, cellular networks can support a diverse range of IoT applications – ensuring the lowest possible TCO.
New downsized NB-IoT chipsets, designed for MTC, and features that improve both coverage and device battery life will boost the ability of LTE infrastructure to address the IoT market. One network that supports all applications – from advanced mobile broadband services, VoIP and all kinds of low- to high-end IoT use cases – creates a very strong value proposition.
Cloud and edge computing serve as the new backbone in the IoT pipeline. As IaaS and PaaS providers offer user-friendly methods to develop and test application without major capital expense on in-house servers, IT support, and or backups/installation. These are key in rapidly and cost-efficiently creating user-friendly applications to unburden IoT from the misconceptions of inflexibility and stalemate, but rather frame it in a way to augment industries and humans in an unforeseen way.
 Ericsson, Ericsson Mobility Report, November 2015, available at: http://www.ericsson.com/res/docs/2015/mobility-report/ericsson-mobility-report-nov-2015.pdf
 Gartner, Gartner Says the Internet of Things Installed Base Will Grow to 26 Billion Units By 2020, December 2013, available at: http://www.gartner.com/newsroom/id/2636073
 Ericsson, Cellular Networks for Massive IoT, January 2016, available at: https://www.ericsson.com/assets/local/publications/white-papers/wp_iot.pdf
 3GPP, Release 13, August 2015, available at: http://www.3gpp.org/release-13
 Medium, True Cloud Story About: IaaS, PaaS & SaaS, February 2016, available at: https://medium.com/@Albihany/true-cloud-story-about-iaas-paas-saas-47cfea883271
 Wolfgang Sommergut, Was Sie über die Cloud wissen müssen, Computerwoche, July 2015, available at: https://www.computerwoche.de/a/was-sie-ueber-die-cloud-wissen-muessen,2504589,2
Microsoft, Internet of Things (IoT) security architecture, September 2018, available at: https://docs.microsoft.com/en-us/azure/iot-fundamentals/iot-security-architecture
 Talari, Edge Computing, January 2019, available at: https://www.talari.com/glossary_faq/edge-computing/
 3GPP, Kevin Flynn, 3GPP on track to 5G, June 2017, available at: http://www.3gpp.org/news-events/3gpp-news/1787-ontrack_5g
 Andreas Knoll, Markt & Technik, Edge Controller statt SPS?, November 2016, available at: https://www.elektroniknet.de/markt-technik/industrie-40-iot/edge-controller-statt-sps-135441.html