Open-source columnar database for real-time analytics on event data
ClickHouse is what you reach for when a dashboard must answer in under a second over billions of events: it is the engine under PostHog, Cloudflare analytics and many ad-tech stacks. Self-hosting is free and a single machine goes a long way. ClickHouse Cloud removes the operations work and bills per compute unit and per TB stored, with an AWS Mumbai region. It is not a general-purpose database; updates and deletes are awkward by design.
Are we wrong about your product? Tell us.
ClickHouse is an open-source column-oriented database built for analytical queries: counts, aggregates and funnels over very large tables, returned fast. Data is stored by column and compressed, so a query touching three columns of a hundred-column events table reads only those three. It is licensed Apache 2.0 and the code is on GitHub.
You can run it yourself (one Docker container is enough to start) or use ClickHouse Cloud, the company's managed service on AWS, GCP and Azure. The Cloud has Basic, Scale and Enterprise tiers; the public rate card shows Enterprise compute and storage prices per region, including AWS ap-south-1 (Mumbai).
Quick facts: Apache 2.0 licence · Columnar, SQL · Self-host or ClickHouse Cloud · AWS Mumbai region in Cloud · Used underneath PostHog and Cloudflare analytics.
Columnar storage, vectorised execution and aggressive compression make GROUP BY over billions of rows routine on modest hardware.
Partition by month, order by the columns you filter on, and set a TTL so old events expire without a cleanup job.
Pre-aggregate on insert (daily active users, per-feature counts) so dashboards read tiny tables instead of raw events.
Ingest straight from Kafka topics and query Parquet files in S3 or Spaces without loading them first.
Window functions, approximate uniques, funnel and retention functions (windowFunnel, retention) built in.
Managed, auto-scaling, with separate compute and object storage; the $300 trial credit lasts 30 days with no card.
Self-hosting costs only your servers. ClickHouse Cloud is usage-based with no monthly minimum; the public rate card lists Enterprise-tier rates per region. Basic and Scale rates are shown in the calculator, not on the page.
| Plan | Price | What you get |
|---|---|---|
| Self-hosted (open source) | $0 | Apache 2.0; you run and back it up yourself |
| Cloud trial | $300 credit for 30 days | No credit card; trial ends at 30 days or when credits are used |
| Cloud compute, Enterprise (AWS Mumbai, ap-south-1) | $0.42138 per compute-unit-hour | 1 unit = 8 GiB RAM, 2 vCPU; US regions from $0.39030 (AWS) and $0.32727 (GCP) |
| Cloud storage, Enterprise (AWS Mumbai) | $27.50 per TB per month | Compressed size; US regions from $22.00 (GCP) and $25.30 (AWS) |
Source: clickhouse.com/pricing, checked 22 Oct 2026. Prices in US dollars; the Mumbai rate is the ap-south-1 row on the rate card. GST applies if you are invoiced in India.
In the order a small team would do them.
docker run clickhouse/clickhouse-server gets a local instance; the Cloud trial gives $300 of credit for 30 days without a card.
MergeTree engine, PARTITION BY toYYYYMM(event_date), ORDER BY (user_id, event_time). The ORDER BY is your index; get it right first.
TTL event_date + INTERVAL 13 MONTH keeps storage flat and makes the retention decision explicit.
Daily active users and per-feature counts as aggregating views, so Metabase or Grafana never scans raw rows.
Use BACKUP TO S3 (works with DigitalOcean Spaces) on a nightly cron; test a restore once before you rely on it.
If ClickHouse is not the right fit, these are the pages to read next.
Serverless warehouse with a free 1 TiB of queries a month and native GA4 export; nothing to operate.
Choose when: You are on Google Cloud or GA4 and would rather pay per query than run a serverProduct analytics, session replay and feature flags on top of ClickHouse, with a free tier of 1M events a month.
Choose when: You want funnels and replays without writing SQL