Understanding the Benefits of Placing Batch Data under System Management

Batch data is data that is processed on a regular basis, usually as part of a production cycle. A majority of these data sets are sequential data sets that are members of generation data groups.

You can readily migrate batch data to system management, but you need to understand application cycles to properly define the availability requirements of the data sets.

You can realize these benefits by migrating batch data to system management, as follows:

You can improve batch job performance for I/O-bound jobs with sequential and VSAM data striping.
Jobs that process large, sequential or VSAM data sets can improve performance if you convert these data sets to extended-format data sets that are striped. Sequential and VSAM data striping causes data sets to be written across multiple DASD volumes on unique storage paths and then read in parallel on each volume.

Channel load and storage space can also be reduced if you use host-based data compression.
You can improve batch job performance by moving tape data sets to DASD.
Batch jobs constrained by tape access speed or drive availability can benefit from system management. Sequential access speed of striped data sets on DASD is faster than that of tape, and you spend no time with tape mounts. Production data sets and backups of production data sets both benefit from system management:
- Production data sets
  Migrating this data to system management gives you an opportunity to improve performance for batch applications by moving selected data sets to DASD.
- Backup copies of production data sets
  Data sets that are application point-in-time backups can be written to a system-managed DASD buffer and managed with DFSMShsm according to the application's space and availability requirements. This strategy is discussed in Optimizing Tape Usage.
Critical batch data sets benefit from dual copy and from RAID technology.
Unrecoverable I/O errors can cause reruns. A cache-capable 3990 model storage control's dual copy capability can diminish the effect of hardware outages by maintaining a secondary copy of the data set on a separate volume. An I/O error on the primary volume causes the system to switch automatically to the secondary copy without any disruption to the application. If you use DASD fast write with dual copy, there is no performance penalty for this increased availability because the I/O is complete when written to the cache-capable 3990 storage control's non-volatile storage. The update is then destaged to the primary and secondary volumes.
Batch data sets benefit from DFSMShsm's fully automatic availability management.
Most applications' availability requirements are met with the STANDARD management class. Applications with specialized cycles, such as batch data sets, might require unique management classes. Using data set-oriented management classes lets you customize services, based on each application's requirements. Later, you can simplify or eliminate application-initiated backup and recovery procedures.

SMS allocates batch data in the PRIMExx and LARGExx storage groups, based on data set size. Data sets larger than 285 MB are directed to the LARGExx storage group.

Most data sets are allocated using system-determined block size to optimize space usage. SMS allocates large, sequential batch data sets having high-performance requirements in extended format. Critical data sets are maintained on dual copy volumes.

For sequential data sets, SMS writes a hardware EOF at the beginning of the data set at initial allocation. This prevents data integrity problems when applications try to read the data before data is written in the data set.

You can manage the majority of your data with the STANDARD management class. Data sets having unique management requirements are identified by data set name or RACF® &APPLIC and managed using a specialized management class. Generation data sets are identified by the ACS variable, &DSTYPE, and receive special management, based on the nature of the generation data group. If the generation data group contains backups of data sets, the current copy migrates quickly to migration level 1. In contrast, if it represents the current version of the production data set, the current version is retained on primary storage until the next generation is created. Output data sets containing reports are early candidates for movement to migration level 1.

Batch data sets can vary greatly from cycle to cycle. Certain data sets should not have space released automatically because of this variability. However, batch data sets that are generation data sets are assigned a management class causing unused space to be released automatically when the data set is closed.