Building a PostgreSQL data warehouse
summary
Describes a production system that turns Postgres into a lakehouse by adding extensions that write Iceberg metadata and Parquet files to blob storage, use DuckDB for vectorized scans, and add logical replication to Iceberg. Shows ClickBench results of ~50–60x speedups, TPC‑H ~14x, a demo moving a billion rows where a 4‑minute OLAP query becomes ~300 ms, and sustained micro‑batches with ~10k writes/sec plus automatic compaction every 10 minutes.
AI-written summary · about
chapters (7)
- 0:00Introduction
- 1:50Differences between OLTP & OLAP
- 6:39Hybrid OLTP / OLAP Architecture
- 9:01Hybrid Table Storage
- 12:54Crunchy Data Warehouse
- 23:30DEMO – Logical Replication from Postgres to Iceberg
- 24:56Summary
related talks
25mpg_lake: Postgres as a lakehouse
34mAnalytics and AI with PostgreSQL - Mission Possible!
19mFrom OLTP to OLAP: Is PostgreSQL Eating Analytics Too?
05mMaking Iceberg walk and talk like Postgres
34mAnalytics and AI with PostgreSQL - Mission Possible!
58mSuper Jumbo Delux
21mAnalytics in Postgres –A decade in the making
25mData-intensive PostgreSQL: Three ways to scale