If the file is thousands of megabytes, or millions of megabytes, you will need to parallelize the backup of this file, usually, a channel can read only one file, but with a multi-segment keyword can change this behavior:
Run {
Allocate channel T1 type SBT;
Allocate channel T2 type SBT;
Allocate channel t3 type SBT;
Allocate channel T4 type SBT;
Backup as compressed backupset datafile section size 10g;
}
This script will start 4 channels, each reading a 10GB segment of the data file 16. Each channel generates slices (separate physical files) that contain the backup of the segment. If the file is 200GB, there will be 20 such slices, generated 4 at a time. Without the section size keyword, parallel reads will be 1 (that is, serialization), with one channel performing the entire operation.
Oracle Rman backs up concurrent executions of individual files