Time-series database
A time-series database stores data points indexed by time.
The core idea is that time is the primary dimension. The database is optimized for appending new measurements and querying ranges of time.
Example
| timestamp | metric | value |
|---|---|---|
| 2026-05-14T10:00:00Z | cpu_usage | 0.72 |
| 2026-05-14T10:01:00Z | cpu_usage | 0.69 |
| 2026-05-14T10:02:00Z | cpu_usage | 0.81 |
Typical queries ask for aggregates over time windows.
What it is good for
Time-series databases are useful for:
- metrics,
- logs,
- IoT measurements,
- financial ticks,
- monitoring dashboards,
- event streams ordered by time.
Tradeoff
They are strong for append-heavy, time-windowed data, but they are not general purpose stores for arbitrary relational queries.
Theory that matters here
Time-series systems revolve around append-heavy writes, retention, downsampling, windowed aggregation, labels/tags, cardinality control, and compression over time-ordered data.
Implementations
- Prometheus — observability system built around a time-series data model.