Automatic Tuning of Read-Time Tolerances for Optimized On-Demand Data-Streaming from Sensor Nodes
Julius Hülsmann, Chiao-Yun Li, Jonas Traub, Volker Markl
Proceedings of the 24th International Conference on Extending Database Technology (EDBT 2021) | March 2021


The Internet of Things provides applications with data streams from billions of sensor devices in real-time. Usually, sensor devices serve multiple queries simultaneously despite having limited computational capabilities. This paper presents a solution for reducing the number of data reads and transmissions by increasing the potential for sharing reads among concurrent streaming queries. Existing read-scheduling techniques on sensor nodes dynamically adjust the data-acquisition rate depending on the data’s variability. However, they leave the definition of read-time tolerances to the user. Such read-time tolerances are crucial for sharing reads among queries. We extend previous work by presenting a generally-applicable algorithm that defines read-time tolerances and adapts them on-the-fly depending on observed data characteristics. We evaluate our solution on real-world data and show that it reduces the sensing error by up to 60% compared to existing approaches with the same number of data reads. Respectively, our technique reduces the number of data reads to achieve the same sensing error as existing techniques. To the best of our knowledge, we are the first to automatically set and tune read-time tolerances to reduce sensor readings and data transmissions on sensor nodes.