Optimizing Tape Usage

In an average installation, 50% of tape data sets have inactive data—data written once and never read again. Most inactive data sets are point-in-time backups. These are application-initiated data set backups used only if an application or system failure occurs. Although data set backup is critical, application-initiated backups to tape make poor use of the tape media and subsystem, and require many costly operator tape mounts. The remaining tape data sets, active data, cause the same inefficiencies in tape cartridge use. However, they are usually not very active; 60% - 90% of all tape data sets have all accesses on the same calendar date.

You can implement these tape mount management techniques without changing your JCL streams or backup and recovery procedures. Tape data sets that cannot be addressed by tape mount management techniques can also be system-managed. Refer to Managing Tape Volumes for information on these data sets.

You can improve tape management efficiency by implementing tape mount management. Tape mount management is a methodology in which you use your ACS routines to re-route tape data set allocations to DASD. Once the data sets are on DASD, you can use DFSMShsm to migrate the data sets to tape as a group, reducing the number of tape mounts required. See Figure 1 for an illustration of how the tape mount management methodology accomplishes this.

The DASD buffer is a staging area for these tape mount management candidates before they are written to tape. DFSMShsm periodically checks the occupancy of the DASD buffer's storage group. If the allocated space exceeds the midway point between low and high threshold (if you specified interval migration) for the storage group, DFSMShsm moves data sets to its migration level 1 or 2 volumes to bring the buffer down to the low threshold.

Data set movement through the storage hierarchy is based on the management class that you assign to the data set. DFSMShsm uses single-file format and data compaction technologies to create a full cartridge for each migration level 2 tape volume before requesting another.

Very large and offsite tape data sets that must remain in their current form are written directly to tape. You can also write them in a compacted form to make better use of both the media and cartridge subsystem.

DFSMShsm automates the management of these data sets by checking the DASD buffer to ensure that space is available when needed, and that data sets are migrated to tape when no longer required on DASD storage. Your tape library inventory is reduced because DFSMShsm tries to fill tapes by stacking data sets in single-file format, in contrast to applications that create separate tape volumes that might only partially fill a tape volume.

Data sets are recalled by DFSMShsm automatically, if accessed by the application. Batch turnaround time is improved by several hours, because jobs are no longer artificially queued up on tape drive availability.