Transparent (and open source) PostgreSQL data tiering: pgEdge ColdFront intro & demo

pgEdge ColdFront: transparent data tiering for PostgreSQL. Unify your OLTP, analytics, and AI workloads on the same connection, with 90% cheaper storage for archival data (and write directly to cold storage).

Most Postgres databases accumulate years of history that nobody queries but everybody pays SSD prices to store. Deleting it is risky. Exporting to flat files makes it unreachable.

Proprietary tiering systems solve the storage problem while creating a new one: vendor lock-in for as long as you need the data, which could be a decade.

pgEdge ColdFront takes a different approach. Hot data stays in native Postgres heap partitions. Cold data moves automatically to Apache Iceberg on S3 - but the application sees one table, uses one connection, and runs standard SQL. Reads and writes work across both tiers. No application changes, no glue code, no sidecar processes.

Paul Rothrock (Solutions Engineer) walks through the architecture and a full Docker demo: setup with SeaweedFS and Lakekeeper, creating tiered and Iceberg-only tables, and the archiver configuration that moves rows automatically on a cron schedule.

Key details:
-> Writable cold tier: UPDATE and DELETE on archived rows work through standard SQL - no restore cycle required
-> Architecture: C extension routes queries, pg_duckdb handles cold-tier reads in-process, a small Go binary runs the archiver via cron, Lakekeeper manages the Iceberg catalog
-> Three modes: tiered (hot + cold), Iceberg-only (full analytic tables), and partition-only for tables not ready for Iceberg
-> Runs on stock unpatched PostgreSQL 16 and above
-> 100% open source under the PostgreSQL License

👉 Check it out on GitHub: https://github.com/pgEdge/ColdFront

👉 Learn more on our website: https://www.pgedge.com/solutions/postgres-tiered-storage