Amazing overview of the various compression schemes used in columnar stores.

I use Parquet with ZSTD with DuckDB and it absolutely flies.

Have not reached for Spark in a long while.