Parallel partition processing

Using selective partition processing, different instances of your application programs can process different partitions of a database in parallel in both the batch environment and the online environment.

DBRC authorizes application programs to process individual partitions rather than the entire HALDB database. Processing partitions in parallel can reduce the time that is required for the processing.

In the batch environment, batch application programs are authorized to process individual partitions one at a time, rather than the entire HALDB database. IRLM is not required.

In the online environment, multiple dependent regions can process records in the same or in different partitions. Data sharing is not required.

If you use block-level data sharing, you can easily process different partitions in parallel with multiple subsystems. The subsystems can be online systems or batch jobs. To use block-level data sharing, you must use IRLM and you must register the databases in DBRC as allowing block level data sharing. For more information about block-level data sharing, see the DBRC information and the data sharing information in IMS Version 15.3 System Administration.

To enable multiple instances of an application program to process partitions in parallel, restrict the database PCB of each instance to the partitions by specifying in a DFSHALDB DD statement the partition name and either the label name of the database PCB or the nth position of the database PCB.

You might also need to make one or more of the following modifications to the input, output, or processing logic of the application program:

• Split the application program input to feed multiple instances of the application program.
• Consolidate the output of multiple instances of the application program.
• Modify the application program to respond correctly to the unavailability of the partitions it cannot access.

For more information about the DFSHALDB DD statement, see IMS Version 15.3 System Definition.