To provide ACID transactional guarantees and dependable version history for files in a data lake is the correct option.

The Delta Lake storage layer implements a transaction log and metadata layer on top of object storage to provide ACID transactions. This means commits are atomic and readers see consistent snapshots even during concurrent writes, and the platform keeps a dependable history of file versions for time travel and reproducibility.

The layer also supports schema enforcement and evolution which helps prevent or detect bad data during write operations and makes it easier to roll back or reproduce past table states for analytics and ETL workflows.

Vertex AI Model Registry is incorrect because that service manages machine learning models and their metadata rather than providing transactional file storage or a versioned data lake.

To act as a distinct tier for long term archival storage is incorrect because a delta layer is focused on transactionality and versioning rather than being a low cost cold archive tier.

To automatically tune queries and build indexes for very large datasets is incorrect because automatic query tuning and indexing are responsibilities of query engines or specialized indexing tools rather than the core purpose of a Delta Lake storage layer.

Full AWS Practitioner Certification Question