Time Series Databases

TimescaleDB, InfluxDB, Victoria.

Overview

Time series databases are storage optimised for time-stamped data: metrics, events, telemetry. Picking one is easy; matching the choice to workload cardinality, ingest rate, and existing operational expertise is the discipline.

• TimescaleDB, InfluxDB, Victoria. Three popular choices with distinct trade-offs; pick to match workload, not vendor preference.
• TimescaleDB: Postgres extension. Reuse Postgres expertise and tooling; the right answer when the team already operates Postgres.
• InfluxDB: dedicated TSDB. Purpose-built for time series; the right answer when no Postgres muscle exists and the workload is pure metrics.
• VictoriaMetrics plus per-workload benchmark. Drop-in Prometheus replacement at scale; production-shaped benchmark is the only honest comparison.

The approach

The practical approach: workload-shaped benchmark first, match by existing operational expertise, monitor cardinality continuously, document the rationale per cluster. The team’s discipline produces matched TS storage instead of marketing-driven choice.

• Workload-shaped benchmark. Production cardinality and ingest rate; vendor benchmarks rarely match the team’s actual shape.
• TimescaleDB for Postgres-heavy. Reuse Postgres operational expertise; one fewer database engine to learn.
• VictoriaMetrics for Prometheus. Drop-in Prometheus-compatible at scale; supports the existing query language and dashboards.
• Monitor cardinality plus documented rationale. Per-database cardinality alerted; per-database choice rationale committed for the next review.

Why this compounds

TS database discipline compounds across services. Each correct choice produces ongoing performance; the team’s storage expertise grows; new metric workloads inherit the patterns.

• Better operational fit. Right database for the workload; the team operates fewer engines and learns each one deeply.
• Better cost efficiency. Right database matches budget; the storage cost tracks the workload shape, not vendor list price.
• Better team expertise. Each database teaches TS patterns; the team’s storage muscle grows.
• Institutional knowledge. Each TSDB teaches storage patterns; the next architectural decision is informed.

TS database choice is an infrastructure decision that pays off across years. Nova AI Ops integrates with TS database telemetry, surfaces patterns, and supports the team’s TS workload discipline.