Time-Series Database for Biomedical Analysis

Abstract

The goal of this thesis is to make a time-series database for biomedical analysis and compare this to an earlier relational database. The system was defined in a previous bachelor task, however, with a relational database as the core. In addition, we extended the system functionality to make use of time-series specific functions.

We decided to use InfluxDB for our implementation, and the web application is hosted by a custom made web server implemented in Go. The time-series we are handling are processed data collected from automated external defibrillators (AEDs).

The relational and time-series database systems are compared through a suite of benchmarks. The benchmark suites consists of testing the basic information retrieval, retrieving incidents based on user-specified criteria, and the time-series extensions. The results suggest that the relational database, in most cases, outperforms the time-series database in terms of speed, while the time-series database has far greater ease-of-use when it comes to time-series related queries.

We suggest that a future system would employ both structures, to retain the best qualities of both the time-series and the relational structure.