For the development of an architecture for the Internet of Things (IoT), we present the approach “Streaming and Data Analysis” in today’s article. The accompanying graphic shows in detail how this looks like.
The “streaming and data analysis” approach for an IoT architecture is more flexible than the “platform as a service” approach in terms of the data types used and is also designed for larger amounts of data. Here, too, there is Device / Thing Management (see No. 7), which in this case is anchored in the business logic. This means that administration is always carried out specifically for a company in accordance with the framework conditions.
For example, not every property of a Thing has to be registered, but only the IDs of the devices and certificates are managed for a secure exchange, while the evaluation then takes place via streams or Message Broker (see no. 8). This can also be an MQTT-based broker, as offered by the IoT platforms explained in the first approach, or a so-called data pipeline, such as Amazon Kinesis Streams. The advantage of these data pipelines is that they can receive and transform any data (and data volumes). In addition, real-time analyses can also be carried out in a SQL-based language (Real Time Analytics, see No. 10), for example to detect anomalies in values.
Even while data is running through the data pipeline, companies detect deviations from rule values and an immediate alarm is triggered even before the data is stored in a warm storage (= data access in the (milli-)second range, see no. 11b) or cold storage (= access speed up to the hourly range if necessary, see no. 11a). Furthermore, the streaming data can be transformed before storage (see no. 9). More complex calculations are also possible, so that important KPIs are calculated almost in real time and stored in the database (see no. 11c).
In a dump truck, for example, the pressure of the hydraulic system is continuously monitored. A gradual drop in pressure in the steering system may not be noticed until damage has occurred. With real-time monitoring, the system would request preventive maintenance of the vehicle if certain data deviated from its set values. Companies can thus prevent cost-intensive vehicle breakdowns in advance by simply analyzing the data.
Once the data has been stored, it can be used for in-depth analyses (see No. 12) with a wide variety of tools. They can also be used for Machine and Deep Learning Frameworks, thus bringing more intelligence to the overall system. These tools also pave the way for predictive maintenance, which optimizes the use of machines and vehicles.
In this second example of an architecture for the Internet of Things, the connection to visualization or business integration (see No. 13) is also made.