Sensor networks are increasingly deployed to monitor and simplify management of physical entities such as planes and wind turbines. As the produced amount of sensor data increases, so do the requirements for methods and systems that can store and analyze the vast quantities of sensor data being collected. However, the systems used in industry are in general not designed to manage sensor data at large scale. This forces practitioners to only store simple aggregates, for example averages over a 10-minute window, instead of the raw time series. As a remedy, this thesis proposes a model-based Time Series Management System (TSMS) named ModelarDB that supports ingestion, storage, and multi-dimensional analysis of time series at scale. A model in this context is any representation from which a time series can be reconstructed within a user-defined error bound (possibly zero). As part of ModelarDB, this thesis proposes methods for model-based management of time series.
All interested parties are welcome. After the defense the department will be hosting a small reception in cluster 3.