Term

Time-Series Database (TSDB)

A time-series database is a database engine built for timestamped sequential data: metrics, telemetry, sensor readings, financial ticks, log events. Writes are heavy, append-only, and arrive in time order. Reads are dominated by time-bucket aggregations (averages, percentiles, rates), not by ad-hoc joins. A TSDB bakes that pattern into its storage layout, indexes, and operational primitives (retention policies, downsampling, continuous aggregates). Representative engines: InfluxDB, TimescaleDB, Prometheus, QuestDB, Apache IoTDB.
Reviewed by 7wData

Why it matters

Most observability stacks you have used are TSDBs under the hood. Prometheus plus Grafana is the canonical pair, and Prometheus is a TSDB. For AI work the relevance is direct: training data for time-series forecasting, anomaly detection, and ops-monitoring models lives in a TSDB before it goes anywhere else. For agentic AI it is sharper. Every action an agent takes is a timestamped event, and the TSDB is where the audit trail accumulates. If you cannot reconstruct what an agent did at minute T-42, you have logs and hope, not observability.

Where you’ll encounter it

Three contexts. A SRE team picks between Prometheus plus Thanos, InfluxDB, and TimescaleDB, and the choice hinges on cardinality limits and SQL familiarity. An IoT product needs to ingest a million events per second, the relational DB folds under the write pressure, and a TSDB is the answer. A model-monitoring stack tracks prediction drift and latency over time, and again the workload shape selects a TSDB. The trade-off is real: TSDBs are weaker than relational DBs for non-temporal joins, so the choice is about workload shape, not about which engine is better in the abstract.

Part of the 7wData AI Glossary. Tracking how concepts like this move in the expert conversation: daily signals at ins7ghts.com.