Lakehouse
A data lake with database properties bolted on: cheap object storage underneath, ACID transactions and schema on top.
The idea solves a real pain. Data lakes were cheap but became swamps — no transactional guarantees, no schema, no time travel. Data warehouses were orderly but expensive and vendor-locked.
A lakehouse stores Parquet files in S3 and layers metadata over them: Iceberg, Delta Lake or Hudi. Databricks pushed the term from 2020. Separating compute from storage — and keeping the format open — is the entire point.