TSM-Bench Benchmark for Time Series Database Systems

This document is a research article that introduces TSM-Bench, a benchmark specifically designed for evaluating time series database systems (TSDBs) used in monitoring applications. The article outlines the necessity of such a benchmark due to the inadequacies of existing benchmarks in addressing the unique requirements of monitoring applications. It details the contributions of TSM-Bench, which include representative queries derived from a water monitoring use case and a novel data generation method utilizing Generative Adversarial Networks (GAN) and Locality Sensitive Hashing (LSH). The authors conduct extensive experiments to demonstrate how TSM-Bench evaluates the performance of seven leading TSDB systems, providing insights into their capabilities and trade-offs. The paper also discusses the challenges faced in selecting appropriate systems for real-time monitoring and emphasizes the importance of a comprehensive benchmarking framework that can accommodate various performance metrics and realistic workloads. The findings are relevant to a broader range of monitoring applications beyond the initial use case.