The SORTPARn DD statements define units that will be used to perform the parallel reading of the input file. Up to four SORTPARn DD statements may be provided, with a minimum of two required. The number of SORTPARn DD statements that you provide may be limited by the tape channel capacity at your installation. See Special Channel Separated Esoteric Names for information on how to determine if your choice is limited.
The n in the SORTPARn is replaced with numbers 1 through 4. The numbers must start at 1 and be numbered consecutively.
The required SORTPAR1 must be coded in one of the following two ways, depending on whether the SORTIN data set is cataloged or not.
-
If the SORTIN DD is defined as a single cataloged data set or as a series of concatenated data sets where the first data set of the concatenation is cataloged, then the SORTPAR1 DD must be coded as follows:
-
If the SORTIN DD is defined as a single non-cataloged data set or as a series of concatenated data sets where the first data set of the concatenation is non-cataloged, then the SORTPAR1 DD must be coded as follows:
where the VOL=SER list contains the identical volumes specified on the SORTIN DD specification. If the SORTIN DD is a series of concatenations, the VOL=SER list contains the volumes that comprise the first data set in the concatenation.
The xxxx is a unit type or generic name compatible with the device associated with SORTIN. If special channel separated esoteric names have been made available, see Special Channel Separated Esoteric Names.
The y is either 1 or 2 and indicates the number of units to be allocated for these devices. For optimal performance, data sets that reside on tapes that can be read only in a single direction, such as 3480s, should have two units allocated. If the data set is on a tape that supports bidirectional processing, a single unit is sufficient. In all cases DEFER mounting must be specified.
The number of SORTPARn data sets to allocate depends on several factors:
-
The total number of volumes to be read from SORTIN and its concatenations.
There is no need to allocate more SORTPARn data sets than total volumes in the SORTIN file. Note that if more SORTPARn data sets are allocated than there are volumes, the excess SORTPARn data sets will be deallocated at PARASORT’s initiation.
-
The degree of performance improvement desired.
Typically, two SORTPARn data sets will provide up to 20% elapsed time improvement; three, up to 25%; and four, up to 33%.
-
The degree of channel contention, which may reduce the number of SORTPARn DD statements used.
The use of special esoteric unit names will ensure that this contention is eliminated, but your choice for the number of SORTPARn DD statements may be limited.
-
Resource availability.
System constraints may limit the number available to a particular job.