When loading large data files with Snowpipe, which approach is recommended to optimize parallelism?

• A. Use a dedicated large warehouse for the 1 GB files.
• B. Merge the smaller 200 MB files into a single file.
• C. Load the files as-is without splitting.
• D. Split the 1 GB files into smaller file sizes.

Answer: (D)

Maximizing parallelism in Snowpipe comes from increasing the number of files it can ingest concurrently. Splitting a large 1 GB file into smaller chunks lets multiple ingestion processes run at the same time, boosting overall load throughput. If you merge into a single file or load as-is, you’re effectively serializing work to fewer ingest tasks, which underutilizes available parallelism. While a bigger warehouse can give you more compute power, it doesn’t inherently raise the number of parallel ingestion tasks Snowpipe can perform. So, breaking the 1 GB files into smaller sizes is the way to optimize parallelism.