CSV vs ORC vs Parquet

adriantarau

Which format in Vertica is better for loading a lot of data (>100 billion rows per day, average file size ~20MB, 50-100 columns)?

Regardless of how long it takes the client to write the file, I'm interested in knowing the fastest way to get the data.

Vertica_Curtis

Parquet. It's easy to use, and faster than Orc. I'm not sure I've seen comparisons between Parquet and CSV, but it's easier to load Parquet generally.