Skip to main content
Version: 0.1

Welcome 🥮

pg_mooncake is a Postgres extension that adds columnstore tables and DuckDB for fast analytics.

Postgres with pg_mooncake ranks among the fastest analytic databases on Clickbench. The extension is maintained by Mooncake Labs.

The Problem​

Postgres is not designed for analytics — it lacks columnar storage and a vectorized execution engine. This leaves Postgres users with the following choices:

  1. Adopt a separate analytics system (e.g., ClickHouse, Elasticsearch)

    This involves complex data modeling, creates challenges when querying transactional and analytical data together, and introduces inconsistencies between systems.

  2. Adopt a Postgres native offering

    While solutions like Citus & Timescale add columnar storage, they lack a vectorized execution engine to fully leverage the format. Meanwhile, pg_duckdb cannot transactionally write data into columnar format.

pg_mooncake adds columnstore tables to Postgres with both storage and compute capabilities that can leverage the columnar format.

pg_mooncake_venn

Why you'll love pg_mooncake​

  1. Transactional inserts, updates, and deletes into columnstore tables

  2. Modern columnstore built on object storage — no disks to manage

  3. Data written as Iceberg / Delta Lake tables (parquet + metadata)

  4. Seamless joins between transactional and columnstore tables

Grab your Slice​