The issue with using a disk journal file is that certain operating systems do not support the read back facility on disks. For ROLLFORWARD, this is solved by running with the SORTED option. The sorted option will only apply the last dbkey occurrence found that is appropriate for the recovery, hence no reading backwards.
Running with the sorted option is usually quicker.
For ROLLBACK, there is a special option that needs to be used with a disk journal. The following is copied from the Utilities manual, ROLLBACK section:
If the journal file is on disk, it cannot be read backwards in certain operating systems. To accommodate this, a SYSIDMS PARM has been created, ROLLBACK3490. To utilize this feature, the journal file must be sorted on the first UTC timestamp of each journal block, in a descending order. In this manner, the ROLLBACK utility, using the ROLLBACK3490 feature, reads each block forward, but is actually getting the blocks in descending order. Although the blocks are in descending order, the journal records within the blocks are in ascending order. As ROLLBACK reads each block in descending order, it processes each journal block from the end to the beginning. So even though the journal file is a sequential file on disk, sorted descending, ROLLBACK is processing the journal records in a descending order. To sort the journal blocks in descending order, use the following sort parm:
? In CMS, you do not need to use the ROLLBACK3490 parameter, but you must presort the journal images as described previously regardless of whether the journal files reside on disk or tape.
? If you decide to use the ROLLBACK3490 feature, or are using CMS, then you must sort the file in descending order prior to running the ROLLBACK utility, even if you are using 'sorted mode.'