AWS Redshift can directly query Parquet files from S3, but I found that functionality to be somewhat poorly documented. It's not clear which Parquet types map to which Redshift types.