When implementing a data pipeline that requires updates, deletions, and merges into a large dataset stored on a distributed file system, what feature should a data engineer use to ensure these operations are performed reliably and consistently, with the ability to handle concurrent modifications without data corruption or loss?