The Internet of Things (IoT) is a driver of digitalization and offers many advantages for businesses. Sensors on networked devices continuously collect data. Evaluating data with a specific goal increases transparency and can optimize production processes as well as develop new services.
Each IoT project can be schematically divided into five phases: collect, transport, store, analyze and archive.
1. The first phase, “collect,” involves capturing sensor data and making it transportable. So-called “retrofitting” plays an important role in the process. It addresses the question: How do I connect older machines to the Internet at all? Production lines have often grown over many years and appliances can be up to 15 or 20 years old. They have internal sensors that output data in milliamperes, for example. This data must then be converted into IP-capable information. There are various approaches. For example, a so-called edge gateway, which translates and then forwards the data. Another method is to use systems that film lamp or switch signals using a webcam, translate them into IP information using artificial intelligence and then forward them for evaluation.
2. The second phase, “transport,” involves the secure and reliable transfer of data from production machines or devices to the data center. Switching, routing, wireless and firewall technologies, for example, play important roles. It is also about selecting the appropriate protocol for communication between the machines. Companies face the problem that there are about 50 different protocols for sensor data communication. Uniform standards are still missing. Only when the devices speak the same language can they communicate with one another.
3. The third phase, “store,” entails storing the sensory data and making it available for analysis. Various technologies are suitable for this purpose, depending on the scenario. High-performance storage solutions, so-called enterprise class storage solutions, are recommended for companies performing data analysis with the distributed system Hadoop. As a rule, SSDs (solid-state drives, non-disk-based storage) are used for storage in industrial PCs for edge computing. Stream analytics, that is, real-time calculations of data streams, always require extremely fast flash resources. Cloud storage is a helpful option for storing large amounts of data in the central data lake. The reason: cloud storage can be scaled very well. It is also important to have a comfortable data management operating system such as NetApp’s storage and data management technology that makes it possible to move or mirror data easily between different storage solutions.
4. The fourth phase “analyze,” comprises the analysis of the sensor data. Here, too, it is important to select the right solutions based on the use case. The framework Hadoop and NoSQL (Not Only SQL) database solutions such as Couch base, MongoDB or Cassandra are suitable for processing large quantities of structured and unstructured data from the data lake. SAP HANA or SAP Business Objects are used for real-time analyses. Many companies also would like to link the analysis results with their ERP system. This link also plays an important role in the “Analyze” phase.
5. The fifth phase, “Archive” is about the cost-efficient, long-term archiving of the sensor data. An important aspect is rule-based, automated data classification. This allows, for example, the system to automatically delete data according to the legal retention period or grant access to the so-called storage tiering. This is a method that distributes data according to its access on different storage media. Thus, data in the initial phase of archiving often first land on faster systems since they are retrieved even more frequently. Later, they can be moved to less powerful, less expensive storage solutions. Outsourcing data into a cloud or withdrawing them from the cloud are also aspects of the archiving phase.
The five-phase model using the example of a networked coffee machine
Is there a time of day when a certain type of coffee is more popular than other times? How much coffee is needed when? And when does my coffee machine need maintenance? Such questions can be answered using the example IoT project, “networked coffee machine.” Innately, a coffee machine has a number of different built-in sensors that measure, for example, how many revolutions the grinder performs, the pressure or the temperature. In order to make the device Internet-ready a small, cloud-based microcontroller which reads out the sensor data and sends it to a Hadoop system via the MQTT protocol could be connected. It saves and analyzes the data. Simultaneously, the analyzed data are matched here with the master data from the ERP (Enterprise Resource Planning) system. This means that it is always possible to access how many revolutions the grinder has already completed and when the next maintenance is due. Based on the database queries, it also quickly becomes clear that the users of the coffee machine prefer to drink espresso at lunch time—information that is of interest for the ordering process.
Following evaluation, the data are migrated for long-term archiving. A previously defined set of rules automatically controls where the data are stored. After six weeks, the system deletes unimportant data and transfers the rest to the cloud, where they are stored at lower cost.
These five phases show just how complex IoT projects are, just from an IT perspective. Different interlinked building blocks are required for companies to be successful in their.
By Morne Bekker, Country Manager at NetApp South Africa