Building a PostgreSQL data warehouse

PostgreSQL is designed as an operational database. The planner, executor, storage layer and other architectural aspects are all optimized for running a high rate of transactions.

Most organizations use a variety of other platforms for analytics, but none are as versatile as PostgreSQL. Integrating operational and analytical database systems requires a huge amount of glue code to orchestrate ETL processes, and in the end you can do a lot less, except run faster analytical queries.

Last year, we set out on a journey to turn PostgreSQL into a modern data warehouse. One built using data lake table formats such as Iceberg, easy access to blob storage, high analytics performance, built using extensions. This talk is about what we learned along the way about turning PostgreSQL into a data warehouse.