Skip to content

Why fdl

The promise of Frozen DuckLake

Frozen DuckLake is a read-only data lake pattern built on DuckLake. Place a catalog database and Parquet files on object storage — that's all it takes to create a fully functional data lakehouse.

The name "fdl" comes from this blog post. fdl manages DuckLake catalogs and their data files, and publishes them to object storage so anyone can query with DuckDB.

  • No database server required — just object storage
  • Cost is storage only (reads are free on Cloudflare R2)
  • No complex catalog service like Iceberg or Delta Lake
  • Anyone can query the data with a single ATTACH statement in DuckDB

This means individuals — not just enterprises — can build and publish their own data infrastructure.

The problem: manual management is painful

In practice, building and maintaining a Frozen DuckLake by hand is tedious:

  • Fetching and re-pushing the catalog database requires multiple steps every time
  • You must specify storage locations repeatedly, making consistency hard to maintain
  • Integrating with tools like dbt scatters configuration across different places
  • There is no standard workflow — each project reinvents the process

What fdl automates

fdl manages the entire Frozen DuckLake lifecycle through a single CLI:

fdl init      # Initialize a project
fdl pull      # Fetch catalog from target
fdl run       # Execute pipelines with injected config
fdl push      # Publish to target
  • Eliminates repetitive manual steps for catalog management
  • Works with any pipeline tool via environment variable injection (fdl run)
  • Makes publishing open data as simple as fdl push

The goal is to make data publishing accessible to everyone — not just enterprises with dedicated infrastructure.

Learn more about DuckLake