The question this section answers: How do table formats give a data lake database-like behavior?

A plain data lake is just files in folders. That means no transactions, no safe concurrent writes, and no easy way to update or delete rows. Table formats (Apache Iceberg, Delta Lake, Apache Hudi) add a metadata layer on top of ordinary Parquet files to fix this. Two pages explain the core ideas:

Term: ACID stands for atomic (all-or-nothing), consistent, isolated (readers and writers do not disturb each other) and durable (once committed, it stays).

Which one? The ideas below apply to all three. Check which format your company platform already supports before choosing, and do not mix them on one table.