In this section, we will discuss the I-IoT data flow. We will look at it first from the perspective of its architecture and network, and then move on to considering the strengths and weaknesses of each available data source option. So far in this book, we have not yet seen a situation in which industrial equipment is directly connected to the cloud. This is actually technically possible—there are several pieces of industrial equipment that have this ability, using internet protocols such as MQTT, CoAP, and AMQP. Examples of this kind of equipment include complex packaging systems, Automated Guided Vehicles (AGVs), robotized working cells, or the devices and equipment involved in a fully-automated warehouse. In this section, however, we are interested in industrial plants in which the factory floor is built and managed by thousands of sensors and actuators. In such a scenario, it is truly not possible to connect all these signals directly to the cloud. There are many reasons why this is the case:

• Variety: In industrial plants and factories, the technology used is always a mix of generations. We might find sensors that were installed just a year ago, using a real-time kernel, with self-diagnostic, auto-tuning, and signal-processing capabilities, together with very old sensors with limited or no fieldbus interfaces. It is not feasible to deal with these different technologies at the same time; we would have to implement different ad-hoc connectivity solutions.
• Amount: The factory floor is made up of hundreds of sensors and actuators and the industrial process uses a huge number of microcontrollers and fully automated working cells. All these signals and information are needed to implement predictive and prognostic analytics on the cloud. They are also the basic building blocks to construct digital twins of the machines or processes. It's not possible to establish a direct connection to the cloud for each of these; the variety and amount of them would require too much maintenance and networking to be dealt with.
• Security: Last but not least, cyber security is a big barrier to implementing connectivity in a factory. Devices and equipment that are not designed to be resilient to cyber-attacks will be much more exposed, thereby putting entire islands of automation at risk.

Let's now move on to looking at the industrial data sources and the related data-gathering techniques.