Um estudo comparativo de sistemas de gerenciamento de banco de dados temporais

Abstract

The work addresses the architectural theme of Temporal Database Management Systems, aiming to analyze, discuss, and experimentally and theoretically compare them, given their growing relevance in the Internet of Things scenario. The solutions analyzed were Apache Druid, TDengine, and TimescaleDB, chosen for their relevance in the temporal data landscape according to rankings in the field. In the theoretical comparison, aspects such as partitioning, table schema, and data model were evaluated. For the experimental comparison, an adapted benchmark model as presented by Visperas (2021) was used. As a result, it was observed that among the solutions evaluated in this work, columnar storage stands out, as well as table partitioning rules over its existence to improve search space. In the experimental results, it was concluded that, overall, the correct configuration of the Time-series Databases resulted in similar performance for the presented data volume.