Time-Series Database Benchmarking Framework for Power Measurement Data

Abstract

Time-series data model emerges to capture sequences of temporal data. Time-series data can be found in various data domains such as economics, epidemiology, social sciences, physical science, and so on. Due to its variation of applications, there are many developments of time series database management systems (TSDBs). However, different TSDBs boast their uniqueness and advantage with comparison of other existing database management systems or particularly other versions of TSDBs. The comparison typically depends on the developers of TSDBs and the target applications. Furthermore, there exists no standard time-series database benchmark. In fact, it has been known that different data domains feature different data characteristics that may affect performances of databases and related algorithms. In this paper, we focus solely on the performance testing of TSDBs over power measurement data. We pay close attention to the characteristics of this data domain, and develop a benchmarking framework based on power measurement data. We use this benchmark to compare 3 TSDBs systems: InfluxDB, OpenTSDB and TimescaleDB as a showcase of our framework. In addition, our framework also considers systems resources in addition to query performances of the databases.