7 Features: IoT platforms must be able to do this

The IoT platform is the heart of the Internet of Things architecture: This is where all data comes together, applications and analyzes are provided and the system is maintained. Due to its fundamental importance, the platform should therefore be selected with particular care. If you consider the following seven points when making your selection, you will come a great deal closer to the goal of a high-performance and fail-safe IoT platform.

There is an unmanageable number of sensors and they all have their respective interfaces and protocols such as COAP, HTTP or MQTT. Battery-based devices that communicate via LoRaWAN and NB-IoT are available to operate sensors independently of electricity and over long distances.

The information from the sensors should be enriched with data from other company applications to enable the best possible insights. So-called message queues, which record the data and transfer it to the IoT platform, are ideal for this. An example of this is the Kafka tool, which separates the data source from the data sink and thus improves the integration of third-party applications.

The incoming information from the sensors must be pre-processed, for which low-code tools are ideal. They have the advantage that non-technical end users can also implement threshold values ​​or calculation logic. It is particularly easy with the visual rule chains from Node-RED. In the open source IoT platform ThingsBoard there is a separate module for this that allows processes to be automated.

Alarm rule chain in Thingsboard
Alarm rule chain in Thingsboard
Photo: it novelty

If the processing is particularly complex and the volume of data is very large, the use of a queuing system is recommended, for example Kafka Streams or KsqlDB.

In addition to processing and storing messages from sensors, IoT platforms must also be able to manage the relationships between devices and other objects. This also includes the maintenance of the metadata (sensor type, location, etc.) and the possibility of being able to map hierarchical structures between the devices. This results in a number of advantages:

  • Dashboards showing all stages from the individual device to the complete structure;
  • user-friendly query options such as “I want the dashboard to show all devices in the large meeting room”;
  • different levels of aggregation for values ​​and alarms, eg “All alarms on the ground floor”;
  • grouping of devices makes them easier to find on the platform;

In some solutions like ThingsBoard it is possible to define multi-layer structures, even within individual customers. Companies should definitely consider this possibility, as many sensors quickly lack an overview if no structures have been defined beforehand. If the individual object-sensor relationships are to be evaluated more precisely, it is advisable to use graph databases (e.g. Neo4j).

Role concepts for multiple users or multi-client capability are an essential building block of IoT platforms if users should not be able to access data from other clients. So that there are no conflicts between mandates, separate processing logics are recommended. With concepts like RBAC (Role-based access control) user groups and their rights can be defined within a client. It should be possible to freely define roles with different access to read, write or both.

The most important part of an IoT platform for the end user concerns the visualization of the data. A platform should not only offer dashboards, it should also be possible to monitor real-time data and historical information.

Data visualization in the form of a dashboard
Data visualization in the form of a dashboard
Photo: it novelty

Functions to initiate actions outside of the dashboard, such as controlling machines or sending gateway commands, are also recommended. In order for data to be comparable over time, the platform must support scalable databases.

Most IoT platforms come with some form of analysis functionality. However, if in-depth evaluations are important, you should pay attention to functions for self-service or management reports. This means that forecasts or period comparisons can also be implemented.

Time series forecast
Time series forecast
Photo: it novelty

The same applies to establishing correlations between different sensor types or root-cause evaluations. These are important so that you can react quickly in the event of a fault. While some platforms already bring these functionalities with them, others offer corresponding extensions for them.

Analysis and detection of an anomaly
Analysis and detection of an anomaly
Photo: it novelty

Anyone starting with an IoT platform should start with a proof of concept on a virtual machine (on premises or cloud). This allows you to first test extensively before the platform is set up as a scalable system Kubernetes and Kafka in the second step. This approach is primarily supported by the ThingsBoard platform.

Deployment is possible in the cloud, as SaaS or on your own hardware. During the test phase, it is mostly based on a monolithic architecture, so that all services run on one server instance, but you can then switch to microservices in order to be able to work scalably and fail-safe (mapping by Docker Microservices, orchestration via Kubernetes). (mb)

7 Features: IoT platforms must be able to do this

Leave a Reply

Your email address will not be published.

Scroll to top