What does Data Lake mean?

A data lake is a storage system that takes in data in its original form without first fitting it into a fixed schema. Tables, text files, images, logs, and sensor readings sit side by side there on equal footing, regardless of their source system. Unlike a data warehouse, a data lake only defines structure at the point of reading, not at the point of writing.

Technically, a data lake consists of object storage such as S3 or MinIO plus a catalog that records each file’s origin, timestamp, and format. Query tools such as Trino or Spark read directly from storage and only interpret the structure at query time. The catalog is not an optional extra: without it, within a short time no one can say any longer which file reflects which state.

A data lake pays off wherever many different kinds of data accumulate and it is still open, at the time of storage, what they will be needed for later. It typically holds measurement series from sensor networks, images, expert reports, and log files from ongoing operations. For a fixed reporting scope with stable metrics, a data warehouse remains the simpler choice.

The advantage over processing data on ingestion is that no information is lost prematurely. New questions can be answered against the existing pool of data instead of triggering an additional collection effort. The effort shifts from preprocessing to ongoing maintenance of the catalog.

A data lake stays usable for only as long as its catalog and access rights are maintained. For personal data, purpose limitation and a retention period are also required, both enforced technically rather than described only in a policy. That is how the data pool stays analyzable even after years.