13 May · 1 min read
Challenge: To implement a high loaded backend system for IoT sensor collection and processing (agriculture vehicles) for one of the largest global manufacturers of agriculture machines
Solution: The server-side solution was initially designed as a seamless scalable solution based on on-premise infrastructure
User Group: Farmers
Implementation of a high-loaded backend system for IoT sensor collection and processing (agriculture vehicles) for one of the largest global manufacturers of agriculture machines. The idea of self-driving machines for harvesting requires a lot of data to be collected on the vehicle side together with close to real-time processing of this data on the server-side. In order to support the chain of harvesting and delivery of the seed to the elevator, an always-on-time approach was implemented by using a prediction mechanism, both for delivery tracks and for harvesters.
Harvesters are very complicated, and the number of sensors should be monitored in real-time. Regular GPS data together with device-specific sensors were processed in real-time (one data flow) however some aggregate metrics were processed by schedule (like total track length, fuel consumptions for the given period, etc).
The server-side solution was initially designed as a seamless scalable solution based on on-premise infrastructure. The design includes different data layer, reporting, storage and analytics modules. One of the biggest challenge that was solved flawlessly is the real-time processing of some set of data that was stored in hybrid storage (partially in memory within further persistence to the distributed database). Different layers for data processing were intended for the separation of data flows along with providing data security and manageability.
With a significant amount of sensors on every vehicle together with a huge amount of data (all devices manufactured already have this module installed), analytics of vehicle usage became one of the main parts of the system. Prediction analytics reports (time to failure as an example) could use dozens of terabytes of data and together with historical information is a very powerful tool for optimizing both – vehicle durability and optimization of manufacturing. Such reports were integrated into customers' dashboards in order to reduce vehicle outages and maintain the durability of the device. Moreover, this will also help in the optimization of logistic chains for repair parts and regular service.