A data lake is typically built on object storage, holding device payloads, logs, images and exports as files, often partitioned by date and source. Because nothing is transformed on the way in, it can absorb anything a fleet produces and serves as the archive from which batch processing jobs, data scientists and model training pipelines draw.
The weakness is well known: without data governance, a lake fills with undocumented files nobody can interpret, and the raw history becomes a liability rather than an asset. Cataloging schemas, recording units and firmware versions as metadata, and enforcing a data retention policy are what keep it useful. The IoT data pipeline shows where the lake sits relative to hot storage.
Part of the IoT Glossary. Spotted an error or a missing term? Tell us.