What is a key benefit of micro-partitioning with columnar storage?

• A. Only columns referenced by a query are scanned.
• B. All columns scanned regardless of query.
• C. Columns scanned in random order.
• D. Columns are ignored if not used in a predicate.

Answer: (A)

Micro-partitioning with columnar storage lets Snowflake minimize the work it does by reading only what the query needs. Data is stored by column, and each micro-partition comes with metadata that lets the engine prune away partitions that can’t satisfy the predicates, plus skip loading columns that aren’t required for the result. Because only the columns referenced by the query (including those needed to produce the output) are loaded, I/O is greatly reduced and performance improves, especially on wide tables or selective queries. So the best description is that only the columns referenced by the query are scanned. Columns required for output are included, not just those used in predicates, and there’s no practice of scanning columns in a random order.