Citus adds a distributed SQL layer to PostgreSQL, automatically sharding tables or schemas across multiple worker nodes and parallelizing query execution. It delivers 20‑300× faster transactional and analytical performance while preserving ACID guarantees, foreign‑key constraints, and full compatibility with PostgreSQL extensions, and is offered as a managed Azure Cosmos DB for PostgreSQL service with automated backups and high availability.
Funding
Funding not disclosed
Founders
Product
Problem
Single‑node PostgreSQL deployments hit hard limits on storage, CPU, and I/O, causing slow queries, high latency for real‑time dashboards, and inability to serve thousands of tenants or billions of events. Scaling out traditionally requires custom sharding logic or moving to NoSQL, which sacrifices SQL semantics, joins, and PostgreSQL tooling.
Solution
Citus extends the native PostgreSQL engine with a distributed SQL layer that transparently shards tables or entire schemas across multiple worker nodes. By routing queries to the appropriate shards and parallelizing execution, it delivers 20‑300× faster transactional and analytical performance while preserving full ACID guarantees, foreign‑key constraints, and the rich PostgreSQL ecosystem. Users can start on a single node and add workers on‑the‑fly; the built‑in shard rebalancer redistributes data without downtime. The extension supports both row‑based and schema‑based sharding, columnar compression for cold data, and integrates with common PostgreSQL extensions (e.g., PostGIS, pgvector, HyperLogLog). For teams that prefer a fully managed offering, Citus is available as a managed database service on Azure Cosmos DB for PostgreSQL, with automated backups, HA, and monitoring.
Target Audience
Primary customers are multi‑tenant SaaS providers, real‑time analytics platforms, time‑series/IOT applications, and microservices architectures that require PostgreSQL scalability without abandoning SQL semantics.
Features
- Distributed SQL engine with adaptive query planner that parallelizes SELECT, UPDATE, DELETE, and COPY across shards.
- Schema‑based sharding (no distribution key required) and traditional row‑based sharding for flexible data models.
- Co‑located shards and reference tables to enable fast joins and foreign‑key enforcement across distributed tables.
- Columnar storage and built‑in compression to reduce I/O and storage costs for historical or time‑series data.
- Automatic online shard rebalancer that redistributes data as nodes are added, with zero‑downtime.
- Full support for PostgreSQL 17 and all major extensions (PostGIS, pgvector, pg_partman, etc.).
- Managed Azure service offering pay‑as‑you‑go or reserved capacity, with HA standby replicas, point‑in‑time recovery, and integrated monitoring via Azure Monitor.
- Tenant‑level metrics view (citus_stat_tenants) for isolating noisy neighbors and capacity planning.