Data lakehouses combine the utility of data warehouses with the cost-efficiency of data lakes by unbundling OLAP databases into layered stacks. Unlike monolithic systems like Postgres, lakehouses separate storage and compute across six layers: object storage, compute, raw data formats (Parquet), table management (Iceberg/Delta Lake), catalogs, and query engines (Spark/DuckDB). This architecture enables cheap, performant ACID-compliant analysis of large datasets. Zero-ETL tools are emerging to bridge Postgres and lakehouse systems, including extensions like pg_duckdb, pg_lake, and pg_mooncake that help Postgres users leverage OLAP capabilities without abandoning familiar tooling.
