file_format: text,seq, parquet
dataset: tpch
compression_codec: none,def,gzip,bzip,snap,lzo
compression_type: none,block,record