A modern open table format that stores metadata in a database—not thousands of files. Petabyte scale and 100x faster metadata operations, entirely driven by SQL and managed on MotherDuck.
Spin up a fully managed lakehouse with MotherDuck handling both metadata and storage, or bring your own cloud object storage.
Copy code
--Create a managed Ducklake and bring your own bucket CREATE DATABASE my_ducklake ( TYPE DUCKLAKE, DATA_PATH 's3://mybucket/my_optional_path/' );
A radically simpler architecture
DuckLake collapses the metadata layer to a single database, avoiding the many-small-files problem that slows down other formats.
Iceberg
DuckLake
A guide to the open table format that replaces file-based metadata with SQL databases for a faster, easier lakehouse. Free early access for data engineers and platform teams.
MotherDuck's MCP server turns your AI assistant into a powerful analytics partner. Ask questions in plain English and get instant, verifiable answers from your data.
Traditional lakehouse formats store metadata in thousands of small files on blob storage. Management means complex, error-prone scripting, and an additional catalog server to keep it all organized. DuckLake is just SQL–ACID transactions, snapshotting, and much more.
Finding the right files for a query becomes a quick, indexed SQL lookup instead of traversing file hierarchies on S3. Schema changes, partition pruning, and query planning all benefit from database-native performance.
With file-based formats, small updates create forests of tiny files that require constant compaction. DuckLake optionally inlines small changes directly into the metadata catalog—no new Parquet files for every micro-batch. MotherDuck handles performance tuning, compaction, and garbage collection automatically.
DuckLakes in MotherDuck are managed like any other database. Connect with any DuckDB client to manage your DuckLake, then share with other users in your organization.
Query your data as it existed at any point in time. No special tooling, no separate snapshots to manage—just add a timestamp or version number.
Copy code
-- Query the table as it was one week ago SELECT * FROM customers AT (TIMESTAMP => now() - INTERVAL '1 week'); -- Or query a specific snapshot version SELECT * FROM customers AT (VERSION => 42); -- See all available snapshots SELECT * FROM snapshots();
DuckLake is available now in MotherDuck. Start with a fully managed lakehouse in one command, or bring your own storage while MotherDuck handles the rest.
