Mount improvement with multi-volume, multi-unit allocation

Prologue: I think this is an issue that needs improvement. I can't find a solution that exists in z/OS - if it does please let me know during discussion of the requirement. I'm posting in MVSE because I believe Allocation belongs to MVSE not MVSS, even though this issue is mostly encountered with tape (virtual or real) allocation.

Requirement: when using multiple devices in the UNIT parameter and concatenated datasets, mounts should be done for the consolidated list of volumes (explicitly listed or via the catalog) in the concatenation.

Example of the problem: A job that reads concatenated tape datasets which are usually multi-volume uses UNIT=(,2) to allow mounts and I/O to overlap (and, on virtual tape systems, to allow virtual volume recall to happen while I/O happens). If the first dataset has only one volume, the system will mount that volume and issue a mount for PRIVAT on the second device. The mount for the first volume of the second dataset is delayed until the end of I/O for the first dataset rather than occurring while the first (only) volume of the first dataset is being read. This results in wasted elapsed time. I believe this is true not just for the first dataset but for any dataset where the number of volumes is less than the number of devices.

This problem can show up after migration of large sequential datasets to different media or compression where a dataset used to occupy multiple volumes and now occupies one.

Comments posted:

Tim Hare
Posted: May 03, 2019 11:17 PM
Mount improvement with multi-volume, multi-unit allocation

-------------------------

I should add, I suppose, though it might be being overly obvious, that this improvement is for disposition OLD/SHR, i.e. the input side of things.

I also have not tried this with DEFER, but will soon.

-------------------------

Tim Hare
Posted: May 05, 2019 06:32 PM
Mount improvement with multi-volume, multi-unit allocation

-------------------------

Behavior with DEFER is different. There is no mount of a PRIVATE volume, but the mount of the first volume of the second dataset in concatenation does not happen until after the first dataset has been read. This happens whether UNIT=AFF=dd1 is coded on the second DD, or if the second DD has its own UNIT=(,2,DEFER)

-------------------------

Stan W Abramczyk, Jr
Posted: Jun 10, 2019 04:12 PM
Mount improvement with multi-volume, multi-unit allocation

-------------------------

I'm in agreement , the system should create a consolidated list of the volumes in the concatenation to be mounted in order, and issue mounts for the first 'n' volumes in that list where 'n' is the number of devices in the UNIT parameter.