article --summary
The summary
The piece explains how DuckLake stores lakehouse metadata in an ordinary SQL catalog rather than scattering manifest files across object storage, so queries can locate the right Parquet files directly. Readers learn how features like file compaction and inlining/flushing help with the small-files problem in streaming workloads.
Holanda also describes why Parquet's layout resembles DuckDB's storage, which keeps conversion work low and lets DuckLake run without large clusters. The article additionally covers Amazon's comments on the DuckLabs acquisition and the DuckDB and DuckLake open-source licensing under the MIT license.
Editorial note: This is a summary of an article first published by The New Stack. The full piece is on their site.