Data Lakehouse

A data lakehouse is an architecture that keeps data as files in cheap object storage — the lake part — while an open table format layered on top provides what the lake lacked: ACID transactions, schema enforcement and evolution, and time travel. The result aims to be one system serving both warehouse workloads (reliable BI) and lake workloads (ML, data science) from a single copy of data.

The load-bearing component is the table format and its catalog; the usual organising convention is the medallion architecture. The honest caveat: more moving parts than a turnkey warehouse, which remains simpler when workloads are purely structured BI.

Go deeper: Data Lake vs Lakehouse: What Changed and Which to Use