ClickHouse

ClickHouse is an open-source column-oriented database for real-time analytics and data warehousing, run self-hosted or as usage-billed ClickHouse Cloud.

ClickHouse stores data by column rather than by row, so analytical SQL queries over large event tables return in real time. The project began at Yandex in 2009, and ClickHouse, Inc. has developed it since its incorporation in 2021. The same engine ships as an open-source download that teams run on their own servers and as ClickHouse Cloud, the managed service. In ClickHouse Cloud, compute-compute separation lets several services share one copy of the data, so separate workloads get their own compute without duplicating the data. Incremental materialized views update as rows arrive, while refreshable materialized views re-run on a schedule. A native JSON column type keeps semi-structured data queryable by path. Table functions read Iceberg, Delta Lake, Hudi and Paimon tables where they sit in object storage, and writing back to Iceberg tables is in beta. Row policies and masking policies decide what each user or role can see, and Python user-defined functions run inside the service. The database suits engineering and data teams whose products depend on fast answers over large volumes of event data, from real-time analytics to observability and AI workloads. Teams that want full control install the open-source build and operate it themselves. Teams that would rather not run clusters pick a ClickHouse Cloud plan: Basic for testing ideas, Scale for production, and Enterprise when single sign-on and private regions are required. Organisations with strict data residency rules can place the managed service inside their own Virtual Private Cloud through Bring Your Own Cloud. Ingestion runs through ClickPipes, the managed integration platform, which reads Kafka topics and other sources into tables continuously. The Kafka and S3Queue table engines do the same job inside the database for teams on the self-managed build. From there the query cache serves repeated dashboard queries, and workload scheduling assigns query slots and memory to the workloads a team defines. The same engine also serves data-lake work, querying open table files in object storage in place rather than loading them first.