What is the cause of large differences between INPUT_BYTES_PER_SEC and OUTPUT_BYTES_PER_SEC for back

We are having issues with slower than expected RMAN incremental merge backups. When querying v$rman_backup_job_details (or RC_RMAN_BACKUP_JOB_DETAILS)  the INPUT_BYTES_PER_SEC is as much as twice as fast as the OUTPUT_BYTES_PER_SEC.  What does that indicate? The backup is not using RMAN compression. The backup is writing to NFS filesystems, could this indicate the NFS file system cannot keep up with the RMAN processes writes?

Also, is there any more detailed explanations somewhere on the INPUT_BYTES_PER_SEC, OUTPUT_BYTES_PER_SEC, and COMPRESSION_RATIO parameters and how to use those to gauge and improve performance? All the documents I can find just barely mention those values.