Benchmarking Time-Series Databases for Industrial IoT Applications

Abstract

As Industrial Internet of Things (IIoT) deployments grow, there is increasing demand for time-series databases
(TSDBs) capable of handling high-ingest workloads while supporting real-time querying and analytics. This
paper benchmarks four leading open-source TSDBs—InfluxDB, TimescaleDB, OpenTSDB, and
Prometheus—using synthetic and real-world IIoT data streams. We simulate sensor feeds from a
manufacturing plant, generating millions of readings per hour across hundreds of sensors. Metrics evaluated
include write throughput, query latency, disk storage efficiency, downsampling capabilities, and integration
with visualization tools. InfluxDB delivers the highest ingestion rate (~500,000 points/sec) and excellent
compression, but struggles with complex joins. TimescaleDB, based on PostgreSQL, supports rich SQL queries
and joins but lags slightly in ingest speed. OpenTSDB scales well with HBase backend but requires significant
tuning. Prometheus excels in monitoring workloads but lacks persistent storage and long-term retention. The
study also analyzes indexing strategies, schema flexibility, and scalability under horizontal sharding. Findings
suggest that database choice should align with workload type—InfluxDB and Prometheus for lightweight
telemetry, TimescaleDB for analytical queries, and OpenTSDB for long-term archival. We provide a
deployment checklist and tuning recommendations for IIoT architects aiming to build resilient, scalable, and
responsive monitoring systems. This benchmark serves as a reference for organizations seeking optimal timeseries
storage solutions.