SnowPro Core (COF-C03) · Free practice question 7 of 15
File sizing for COPY INTO
You're loading several gigabytes of CSV files into Snowflake via `COPY INTO` and want the highest load throughput. Which file-size strategy aligns with Snowflake's documented best practice?
- A.One large file of any size — Snowflake parallelizes inside a single file.
- B.Files of approximately 100–250 MB compressed, loaded concurrently.
- C.Many tiny files under 100 KB each, for maximum parallelism.
- D.Exactly one file per target micro-partition size of 16 MB.
Show answer and explanation
Correct answer: B. Files of approximately 100–250 MB compressed, loaded concurrently.
Why: Snowflake's documented sweet spot is files of 100–250 MB compressed because that balances parallelism (the warehouse splits work across files) against per-file overhead. A single huge file limits parallelism; tiny files multiply overhead; sizing to micro-partition size is unrelated to load-side file sizing.
More free SnowPro Core (COF-C03) questions
- Micro-partition immutability
- SECURITYADMIN for user and role management
- Multi-cluster warehouses for concurrency
- Snowflake edition for extended Time Travel
- Recovering a truncated table with Time Travel
- Types of internal stages
- Query result cache
- Reader accounts for non-Snowflake consumers
- LATERAL FLATTEN on VARIANT arrays
- Streams and tasks for change data capture
- Clustering keys for partition pruning
- Network policies for IP allowlisting
- Zero-copy cloning for QA environments
- Snowpipe auto-ingest for low latency