Citus Data

Market position

Citus positions itself as distributed Postgres, contrasting with earlier approaches that forked the underlying database for offline analytics and with architectures that combine non-relational technologies such as Hadoop and NoSQL alongside relational databases. The company described CitusDB as capable of serving both real-time mixed analytic and short-request workloads, and as a distributed database that stays current with the latest version of upstream PostgreSQL.

Citus is delivered as a Postgres extension rather than a fork, so clusters can track the latest PostgreSQL releases and reuse existing Postgres tooling, extensions and ecosystem. It supports both transactional and analytical workloads in one database, requires few application-layer changes and no separate middleware, and offers schema-based sharding so existing applications can be onboarded without choosing a distribution key.

Technology

Citus is a PostgreSQL extension (not a fork) that adds distributed tables and distributed schemas. It provides a distributed query planner and adaptive executor that parallelize SQL across shards and nodes, full SQL query and transaction routing, distributed transactions, reference tables replicated for joins and foreign keys, co-located shards, parallel distributed SELECT and DML, parallelized COPY, and columnar storage for compression and analytics. Operationally it offers an online, non-blocking shard rebalancer (with a background rebalancer added in 11.1), single-node to multi-node scale-out, tenant isolation and monitoring via citus_stat_tenants, cluster activity views (citus_stat_activity, citus_dist_stat_activity), user and access management across the cluster, metadata syncing for query-from-any-node (Citus 11), change data capture in preview (Citus 11.3), and high availability with Patroni 3.0 (Citus 11.2). Citus 12 introduced schema-based sharding; Citus 14.2 supports PostgreSQL 18. Packages are available via Docker, Ubuntu/Debian and Red Hat repositories, and source is on GitHub.

Go-to-market

Distribution combines free open source downloads and GitHub availability with commercial enterprise and managed-service offerings. The company invested in developer and community marketing, including PostgreSQL conference talks (beginning at PGCon Ottawa in May 2015), organizing the first PGConf Silicon Valley in November 2015 in cooperation with the San Francisco PostgreSQL Users Group, and technical blogging. The Postgres team at Microsoft continues to host Postgres events, including the annual POSETTE virtual event, and produces a monthly Postgres podcast. Proceeds of the 2015 Series A were earmarked partly for expanding sales and marketing.

Multi-tenant SaaS applications that can shard by tenant, customer or account identifier; teams building customer-facing real-time analytics dashboards requiring sub-second responses for thousands of concurrent users while ingesting fresh event data; microservices architectures where each service owns a schema in a shared distributed database; and enterprises already using relational databases that need to scale to billions of data points. Time series and IoT workloads are also cited as fits for row-based sharding.

Geography

Headquartered in San Francisco (SoMa district) after moving from Istanbul in March 2013, with continued offices in Istanbul, Turkey, and a third global engineering office opened in Amsterdam in May 2018. During 2015 the workforce spanned the United States and Europe, and the company described a distributed workforce across five countries.

History

Citus Data was founded in January 2011 and joined Y Combinator's Summer 2011 batch. Its first production user, MixRank, went live in July 2012, and the headquarters moved to San Francisco in March 2013. The company released the cstore_fdw columnar store for PostgreSQL in April 2014 and the first open source release of the pg_shard extension in December 2014. In October 2015 it raised a Series A led by Khosla Ventures. In March 2016 Citus was "unforked" and reimplemented as a Postgres extension, and the extension was open sourced; Citus Cloud, a fully managed database-as-a-service, reached general availability in July 2016. A third global engineering office opened in Amsterdam in May 2018, and in November 2018 the company donated 1% of its stock to non-profit PostgreSQL organizations in the US and Europe. Microsoft acquired Citus Data in January 2019. With the Citus 11 release in June 2022, querying from any node became possible and Citus became 100% open source.