The fastest open-source analytical database

ClickHouse uses a sparse primary index - just a few megabytes per terabyte - scaling effortlessly while pruning data quickly. Query caches, advanced skip indices accelerating performance further. Inserts are equally fast, streamed row by row or in batches, with writes optimized through part sorting and separation of reads and writes, delivering throughput without slowing queries.

A parallelized, vectorized engine filters and aggregates data at speed across a single server or hundreds of nodes thanks to partial states, support for sharding and shared processing across replicas.

ClickHouse scales vertically, fully using machine resources and parallelizing reads at a low level.

A decentralized architecture with sharding and replication enables horizontal scaling to hundreds of nodes and quadrillions of rows.

ClickHouse Cloud builds on this foundation with separation of storage and compute: nodes read from a single authoritative copy in object storage, cache data locally, and scale compute dynamically both vertically and horizontally while all reads and writes stay consistent.

ClickHouse offers advanced features to manipulate, filter, and transform data efficiently. Materialized views can be refreshable or incremental, with incremental views shifting compute from query time to insert time and dramatically accelerating repeated queries.

Projections let you maintain the same data sorted in different ways, enabling ClickHouse to optimize for multiple access patterns and deliver consistently fast queries.

A column-oriented design delivers high compression, while the LSM-inspired engine and background merges keep parts compact and queries fast, no matter how large the dataset.

ClickHouse is built on a unique architecture with a pluggable storage layer: data can live on SSDs, spinning disks, or object storage, and can naturally flow across tiers from hot to cold.

ClickHouse ensures reliability through its replication model, which uses Raft-based coordination via Keeper and provides either immediate or eventual consistency depending on requirements. This lightweight design enables replication across availability zones or even regions with high latencies, delivering high durability without heavy overhead.

For long-term protection, backups can be written to object storage, giving teams confidence their data is safe and recoverable.

Snapshots offer a lightweight means to create a point in time of your data.

With support for 70+ file formats for ingestion and output, ClickHouse delivers unmatched interoperability.

The ability to read and write open table and lake formats such as Parquet, Iceberg, and Delta, with catalog integrations like AWS Glue and Unity making them seamless to query - bringing the performance of ClickHouse's query engine to your data lake.

ClickHouse is a full database engine with complete SQL support, including joins, and an optimizer that can reorder joins globally and leverage column statistics automatically. It extends standard SQL with 100s of analytical functions, making complex aggregations and filters simpler and more expressive.