What is the recommended data file size to optimize parallel load operations?

• A. 100-250 MB (or larger) in size compressed.
• B. 1-2 GB
• C. 10-50 MB
• D. 500-1000 MB

Answer: (B)

To optimize parallel load operations, you balance how many files you have with how much data each file contains so many workers can work at once without spending excessive time on overhead from many tiny files. The recommended range of 1-2 GB per compressed file gives enough data per file to keep multiple parallel tasks busy, while still allowing Snowflake to distribute work across several workers efficiently. If the files are too small, you end up with too many tasks and higher per-file overhead, which can reduce overall throughput. If the files are too large, a single file can become a bottleneck and limit parallelism. Smaller options (like 100-250 MB or 10-50 MB) tend to increase overhead and reduce efficiency, while very large files (500-1000 MB) can underutilize the available parallelism.