Data Lake

A data lake is a repository that can retain data in varied forms, often using files in object storage. It may contain original extracts, event records, images, and curated analytical tables. A lake is not limited to unstructured or unprocessed data.

Keep the material, then make it usable

A retailer might preserve order exports and product photographs in the same lake while maintaining cleaned order tables in a separate area. The photographs and tables support different uses, even though they share storage infrastructure.

Keeping the original export can support a recalculation after a business rule changes. That possibility depends on retaining the required data and the information needed to interpret it. A folder of unexplained files is not a reproducible record by itself.

Flexible storage still needs structure

Schema-on-read means applying an interpretation when reading data rather than requiring every incoming source to fit one central model first. It does not mean that records have no structure. File formats, field meanings, versions, and access rules still matter.

AWS describes lakes as accommodating diverse data and analytical uses. To make that flexibility useful, teams need a catalog, ownership, quality checks, and retention decisions.

A warehouse emphasizes managed analytical tables and query workloads. A lakehouse adds table management and analytical capabilities over lake storage. These roles overlap: the word lake does not tell you whether a particular dataset is raw, trustworthy, or ready for a report.

Reference: AWS: Data lakes.