The IoT has gone from a promise for the future to being part of the mainstream set of technologies in both consumer and business market segments. The internet has gone through its evolution over the past several years. It started with connecting computers so they can communicate with each other, which was quickly followed by the phase of connecting people (online social media). Now that all things around us generate data, have some sort of sensors associated with them, and can be connected to the internet, it becomes significant that this third phase of the internet, where all things the generate data can talk to each other is set to grow exponentially in the coming years. This book is targeted toward architects and engineers who intend to build applications for the IoT domain; as connectivity between devices grows, it's applications that will harness the power of interconnectivity and generate useful knowledge to drive productivity for businesses and provide the next level of services for consumers.

We will start by looking at the different parts of the technology stack that make up the IoT, what are the specific parameters required to satisfy the technology solutions, and how they interact and communicate with each other. The intent is to go deeper into these technologies and provide an end-to-end framework for developers to design and build IoT applications for any industrial vertical. We will also provide sample source code for developers to build upon for each component such as the edge, data, analytics, communication, and cloud portions. We will be covering use cases in various verticals as examples to provide some real-world scenarios for readers.

The IoT has already started to change the way large companies operate. A large majority of these companies are currently using the IoT to track their customers, products, and supply chains. Investment in the IoT is therefore going up significantly year after year for such businesses. Some trends in this spending can be forecast based on returns. For example, predicting the operating efficiency or preventing a major outage of a high-priced service offering will result in higher returns, so it's natural that corporations providing such high-priced offerings will invest more in this area. This is evident from the fact that industrial manufacturers that have very high priced large asset product and service offerings are the ones that have reaped the maximum benefits of the IoT when compared to other major industry segments. Mobile and sensor-based tracking of goods through the supply chain and tracking their usage by customers will continue to increase over the next several years.

Today, companies in the United States and Europe are leading the investment in this area but Asia-Pacific has already caught up fast. So far, the largest impact has been in the area of better customer service and faster product improvements for customized offerings. Adoption of these technologies will require the businesses to overcome challenges such as identifying new business models, product offerings, and revenue sources as they adopt these technologies; overcoming cultural shifts that are needed to enable the full adoption of these technologies; identifying what data to collect and analyze and how to secure this huge data stream; and integrating that with existing legacy systems and data.

We think the greatest impact will be mostly felt in the following areas:

• Enhanced servicing and availability of information for products and service offerings
• Increased uptime in services to customers; this will be attributed to reduction in downtime by predictive monitoring and pre-emptive fixes and replacements

• Emergence of new business models in the industrial sector due to better insights and leveraging of data
• New industrial products and service offerings that are more autonomous

We will define three broad areas in which all of the technology pieces required for enabling the IoT can be classified. The first one is edge-related technologies, which enable all of the activities that can be performed close to the source of the data. This includes sensors and sensor hubs that interface with the actual machines and devices. These perform data acquisition, basic data filtering, and data transfer to the cloud or gateway. Data from the sensor hubs flows to controllers or mobile devices that are used for storage and perform advanced data cleaning and filtering. The analytics on the edge happen on the gateways where data aggregation and analytics take place. Edge applications can also be deployed on the gateways, which can perform the real-time analytics to be done close to the source the of data. Apart from these, there are management applications that are required to manage devices and edge analytics applications. A platform to support the runtime environment for these also needs to be provided on this layer. Now, let's check the following diagram:

A key component of the IoT architecture is the gateway, which acts as the interface or proxy between the edge devices and the cloud/enterprise data center (see smart gateway in the preceding diagram). Apart from handling the connectivity between the edge and the cloud, the gateway acts as the interpreter/translator between field and cloud protocols. These gateways are increasingly becoming smart as they are embedded with logic to direct traffic based on rules, such as should data be transmitted as is or does it need to be aggregated or filtered prior to being transmitted? The following table describes the commonly used protocols on the edge layer:

Similarly, we have prepared a table for the common cloud protocols that are used as part of the IoT stack, which is presented as follows:

The second important piece is the Connectivity layer, which connects all edge-related technologies to each other and further to the cloud. Often times, the IoT device network is geographically dispersed and edges require multiple carrier relationships for secure Quality of Service (QoS) enabled networks. This requires policy driven connectivity, security, and QoS for the edge components and additionally policy authoring, management, and deployment on the cloud. Mobile sites require selection of networks based on cost, availability, and bandwidth in a dynamically changing environment.

In order to achieve this, we need a connectivity layer that is aware of IIOT applications and data semantics; a layer that is a Software Defined Network (SDN) and based on Machine to Machine (M2M) and Machine to Client (M2C) communications. Variable network characteristics of M2M and M2C require adaptable traffic shaping based on the nature of applications. Carrier-agnostic, QoS-enabled secure tunnels for M2M and M2C connectivity are required.

The third type of technologies is the cloud analytics platform capabilities that will process and analyze the data coming from the edge devices. The analytical platform provides capabilities to ingest, store, search, analyze, and finally visualize or consume data. The different types of technologies that enable each of these capabilities are required as part of this analytics stack. The requirements for each part of the stack are varied as well. For example, different types of ingestion capabilities, such as real-time ingestion, batch ingestion, and change data capture ingestion are required to be part of the complete stack. Similarly, the storage requirements for different types of data are different. For example, time series data requires NoSQL databases and image data requires Blob storage. A partial list of these capabilities is shown in the Common protocols utilized at the edge table, which gives an idea of the variety of technologies that are required to complete the stack.

As the penetration of the IoT continues in various industry verticals, we will see new business models evolve and new and creative ways of generating value for the end customers and therefore generating revenue. Before we look at possible emerging business models, it's important to take a deeper look at how these technologies are being used today. We can divide the use of IoT technologies into the following broad categories:

• Smart monitoring of the existing product or service install base: The products and services that businesses sell today have monitoring, alerting, and reporting solutions. However, these are siloed and not interconnected with each other or to the internet. That limits their utility and usage; the IoT will enable these monitoring solutions to be smart in the true sense of the word.
• Smart remote diagnostics: It is another area where businesses have started using IoT technologies. This is especially useful for verticals where the operating conditions are harsh and not suitable for human intervention and involvement at all times. The ability to seamlessly integrate edge solutions with the internet has enabled companies to take remote operations to the next level.

• Cross vertical domain data integration: This allows businesses to generate new insights by integrating data and correlating parameters to identify new trends. An example would be the integration of weather data with operational maintenance data for an asset that is deployed in the field and exposed to weather conditions. The integration of these data points will help discover new models for remote monitoring and diagnostics for the asset.
• Product/service promotion: It can be augmented by IoT devices as they can help transmit messages to a customer's smart phone and other devices, which will help them be aware of new promotions and products available in their vicinity. They can also help with targeted ads to be delivered on billboards, and so on.
• Creating open and scalable interfaces for products and services that were earlier closed systems: For example, the automobile industry, which produces engines and power trains for vehicles, are now exposing these assets through interfaces such as software APIs that help with the tracking and maintenance of these systems.

The continuation of these trends will lead to new business models. We predict some of the following and this is by no means the entire list. There are models that will emerge that we cannot comprehend today:

• Revenue sharing models: It is based on the efficiencies created by using IoT and advanced analytics. As an example, General Electric (GE) produces smarter wind turbines for its renewables business. These wind turbines have several sensors and IoT applications that can gather and harness data, combine with weather data, and predict service operations and maintenance opportunities. GE upgrades the existing install base at the customer site for no cost and the costs saved and additional revenue generated by the customers are shared with GE.
• Service ownership models: Existing product companies will package their offerings as services and take ownership of the end to end delivery for the service. So essentially, they are not selling them outright but rather leasing them and moving to a product as a service model.
• Data-driven business models: It will become prevalent as businesses deploying IoT solutions will gather a variety of customer data. So, they essentially become custodians of their customer data and, in partnership with the customer, they can use this data to generate additional revenue by packaging and selling it.
• Efficiencies in the supply chain: It will result in eliminating unnecessary middle layers and hence, generate additional cost optimizations.

Here we describe use cases for three industry verticals that have deployed and started using IoT solutions. The healthcare industry benefits through enhanced patient monitoring capabilities as well as monitoring and management of medical devices, while manufacturing and aviation are trying to address their high availability and quality challenges through IoT solutions. We will provide some details about these in this section to give you a good understanding of these existing use cases.

With the growth in the medical device and equipment market, there is an increased need to manage the growing install base and harness it through analytics for generating new insights for cost reduction, identify new sales opportunities, and respond to customer needs. Optimizations around the supply chain can be accomplished by having a real-time view of the install base.

The use cases for improved install base visibility fall under four categories. The first one is the use case of upgrades, which allow for accurate upgrade offerings enabled by more granular install base visibility aligned to upgrades. Automation of top upgrade campaigns has led to several million dollars in operational productivity.

The second use case is about parts pricing. Parts price getting and price setting algorithms drive margins as well as additional benefits from setting better parts pricing strategies. This requires automation of price getting reports and engagements of SMEs for price setting.

The third use case is about product quality and reliability. The focus in this use case is on system analytics to identify parts and systems that fail more. Early identification can lead to up-selling opportunities as well as troubleshooting to reduce cost to serve. Implementation of a parts recommendation engine helps identify what parts fail more often and why.

Lastly, inventory optimization can lead to significant cost reductions. This is accomplished by driving customer segmentation to fulfill parts to customers based on their entitlement and also managing inventory through better visibility to systems and system failures.

The reduced margins in manufacturing have led companies, such as GE, which are based on heavy industries, to rethink their strategy beyond just automation. The idea expands automation to creating intelligent factories that are not only automated but also continuously learning and improving. This use case will change the way the main personas in this space work and interact—plant managers, manufacturing engineers, plant and business team leads, and quality professionals.

In simple terms, the business use case that will enable manufacturing to become brilliant is centered on the effective use of business intelligence, data science, and self-service models. A proper Business Intelligence (BI) strategy should help identify What happened? and When did it happen? The strategy should incorporate utilization of reports, dashboards, trending KPIs, and related genealogy (traceability) for it to be effective. Proper application of data science advanced statistical analysis, modeling, and machine learning will lead to answers to Why it happened? and What could happen? Finally, data preparation and blending for root cause and data quality analysis is a requirement that should be automated as self-service for effective use.

The production of aviation turbine blades requires data science models to be built in order to identify and optimize the decisions to scrap parts instead of shipping with some impacting defects. This requires applying advanced analytics techniques to help realize the ROI through this improved quality control. Typical blade operations consists of the following steps:

1. Create router and drill EDM holes
2. Grind dovetail and CMM inspect
3. Heat treat (transfer data has run number in it FYI)
4. FPI inspect
5. Visual inspect
6. Farm out transaction to OV
7. Coat at OV
8. Visual inspect
9. Farm in transaction from OV
10. Shotpeen
11. Waterflow

12. Airflow inspect
13. X-ray inspect
14. Final visual inspect and package for shipment
15. Ship transaction
16. Closes router

In a case study, the data engineering team created flattened data by serial number for these steps, which was the framework for all data analysis. This included data such as percent of pieces above and below control limits, by Coordinated Measuring Machine (CMM) dimension. It was identified that the top defects were due to EDM and grind. It was also identified how grind defects correlated to CMM dimensions. This was accomplished by grouping all serial numbers into two groups—those that had grind defects (red) and those that don't have any grind defects (green). The frequency distribution (d) of four key dimensions was plotted to see if there was a perceptible difference in the two groups. See the first figure of the chapter.

The data science effort involved predicting each KPI based on diverse data sources. Analytics models were developed for two KPIs—Part Defect Rate and Machine Uptime. It was identified that mostly EDM and casting issues lead to scrap and some grind defects can be recovered. Plots were generated for the count of serial # with at least one defect, the count of scrapped and shipped, as well as the defect count by type (EDM, cast, and grind). CMM data shows statistical differences in individual measurements for parts shipped healthy with no defects, shipped with grind defects and scrapped with grind defects. See the Common cloud protocols table.

In this chapter, we provided an overview of the entire IoT landscape. We covered the business impact and the technology stack components and described use cases for major verticals. This should have given you a good idea of this entire space, made you familiar with the different pieces of the IoT puzzle, and prepared you for going deeper into some of these topics.

In the next chapter, we will introduce you to how you should consider the overall architecture and design when building out IoT applications. We will take you through the details that are required for you to start visualizing and building your applications.