Skip to main content
Version: 0.2

Welcome 🥮

Preview

pg_mooncake v0.2 is in preview. Expect frequent updates. Join our community for the latest.

pg_mooncake is a Clickhouse alternative for real-time analytics on your Postgres tables.

It allows you to create a columnstore copy of your updating Postgres tables that is transactionally consistent (sub-second latency).

Queries on columnstore tables are executed by DuckDB and rank among the fastest on ClickBench; columnstore data is stored as Iceberg tables (parquet + metadata).

pg_mooncake is maintained by Mooncake Labs. It packages Moonlink, a real-time storage engine for Apache Iceberg, as a Postgres extension.

The Problem​

To run fast analytics queries on Postgres tables, developers typically replicate data into specialized analytics databases like ClickHouse, creating two key challenges:

  1. Complex ETL pipelines and data modeling
  2. Maintaining transactional consistency between OLTP and columnstore. Columnar systems often struggle with concurrent updates/deletes, and often store multiple copies of data.

pg_mooncake creates real-time columnstore copies of your Postgres tables with efficient updates/deletes. No CDC, ETL, or data modeling required.

It is the the first real-time columnstore built into Postgres. Read more about Architecture.

Why You'll Love pg_mooncake​

  1. Analytics on latest transactions: Run OLAP queries on latest Postgres state.
  2. Sub-Second Postgres to Iceberg Sync: Create Iceberg tables that automatically stay synchronized with your Postgres data.
  3. Trivial to get started: One line of SQL to create an Iceberg table that's always in sync with your Postgres data.

Get Started​