I/O buffers are not shared between batch jobs. When multiple jobs concurrently read the same data set, each job reads the data into its own buffers. You can use the Hiperbatch facility to eliminate a separate I/O for each job when multiple jobs accessing the same data are scheduled to run concurrently on the same processor.

For selected data sets, the shared buffer is established on the first open of the data set by any of the jobs. The access methods use the Data Look-aside Facility to cache the data in the shared buffer. For example, the shared buffer is invalidated when another job opens the data set for output.

The candidate data sets for Hiperbatch are defined to the system. For example, physical sequential data sets accessed using QSAM and VSAM ESDS, RRDS, VRRDS, and KSDS with a control interval size of a multiple of 4096 bytes are eligible for Hiperbatch.

Related Reading: For more information about Hiperbatch, see the MVS™ Hiperbatch Guide on the z/OS Internet Library