Created an empty 3350 to test load a AWSTAPE containing one million 164 byte records. Not surprisingly it's a 164MB Linux file. The AWSTAPE created from that file is 164.1MB. I tested with IEBGENER and a COBOL program and each took around a whole 3 seconds to load the data. The CPU time for the COBOL program was slightly less, but not enough to matter. However the COBOL program saved one million bytes because I could control the output and didn’t have to output the line ending character. We were very aware of those extra bytes…back in the day.
For whatever reason this was something I was very curious about. To see how efficiently MVS under Hercules, handled a large amount of records.. Pretty well IMHO.