Time Series Databases

InfluxDB31.7k★

InfluxDB is a time-series database built for metrics, events, and IoT data, handling millions of writes per second with built-in downsampling and retention policies. Regular databases can store time series data, but InfluxDB is optimized for high write throughput and fast range queries like "show me CPU usage for the last 24 hours." Apache 2.0, recently rewritten from Go to Rust (v3). The query language is SQL-compatible now (v3), dropping the custom InfluxQL and Flux languages from earlier versions. Handles millions of writes per second. Built-in downsampling, retention policies, and continuous queries. InfluxDB Cloud has a free tier: limited writes and queries, 30-day retention. Self-hosting is free with no feature restrictions. Cloud paid plans start at ~$0.002/MB written. Self-hosting v3 requires some infrastructure knowledge but isn't awful: single binary, configure storage. The ops burden is moderate: you need to manage retention policies, backups, and storage growth. Time series data accumulates fast. Solo: cloud free tier for prototyping, self-host for production. Small teams: self-host, $10-30/mo in compute. Medium to large: evaluate InfluxDB Cloud vs self-hosted ops cost. The catch: InfluxDB has gone through three major architecture rewrites (Go monolith → Go with TSM → Rust with Apache Arrow). Each version had a different query language and different trade-offs. The ecosystem documentation is fragmented; make sure you're reading v3 docs, not v1 or v2 guides.

TimescaleDB23.3k★

TimescaleDB makes Postgres fast at time-series workloads: server metrics, IoT sensor readings, financial ticks, event logs. It's not a separate database, it's an extension on top of Postgres, so your existing tools, ORMs, and SQL knowledge all still work. The key trick is hypertables, which automatically partition data by time so a query like average CPU over the last hour stays fast even across billions of rows. Continuous aggregates pre-compute rollups, and compression cuts storage by 90%+ on time-series data. The community edition is free and open source to self-host. It installs as a Postgres extension, so if you can install extensions, you can run it. The company behind it rebranded from Timescale to TigerData, and the managed offering is now Tiger Cloud, positioned as fast Postgres generally rather than just time-series. The open-source extension is still called TimescaleDB and still ships under Apache 2.0 plus the Timescale License for the community features, so no license change, just a new name on the cloud. Solo: self-host on your existing Postgres, free. Small teams: self-host, or use Tiger Cloud, which starts around $30/mo for the Performance tier and $36/mo for Scale, with a 30-day trial. Medium to large: weigh Tiger Cloud against self-hosting based on whether you want to own HA and backups. Self-hosting stays straightforward here since it's just an extension. The catch: some advanced features (multi-node, certain compression options) sit under the Timescale License rather than pure Apache, so check it for your use case, especially if you're building a competing managed service. And if your workload is pure metrics with Prometheus-style queries, VictoriaMetrics or InfluxDB may be more purpose-built than general Postgres.

QuestDB17.2k★

QuestDB is built for high-volume time-series data: server metrics, IoT sensor readings, financial ticks, application logs. It's a time-series database that ingests millions of rows per second and runs SQL queries across billions of rows in milliseconds. The performance comes from a column-oriented storage engine written in Java and C++, optimized specifically for time-series patterns. It speaks Postgres wire protocol, so your existing tools (psql, any Postgres client library) work out of the box. Also supports InfluxDB line protocol for ingestion. Competing directly with InfluxDB, TimescaleDB, and ClickHouse in the time-series space. Self-hosted is free under Apache-2.0. No feature gates. Enterprise adds clustering, role-based access, cold storage, and dedicated support, pricing is custom but expect $1,000+/mo. The catch: single-node only in the open source version. If your data outgrows one machine, you need Enterprise for clustering, and that's where costs jump. Also, QuestDB is purpose-built for time-series. Don't try to use it as a general-purpose database. For mixed workloads, TimescaleDB (Postgres extension) gives you time-series performance without giving up Postgres features.