Disk work storage devices

Program performance is improved if you use devices, storage areas, and channels efficiently. If you specify a particular device type with the UNIT parameter on the DD statements that define intermediate storage data sets (for example, UNIT=3390), DFSORT assigns areas, and some optimization occurs automatically. You can get the best performance using disk intermediate storage devices when you:

Use two or more work data sets.
Select the storage device with the fastest data transfer rate.
Assign one work data set per volume if Parallel Access Volumes (PAV) are not used. This will help minimize I/O queue time on the SORTWKdd data sets.
Use devices that are of the same type.
Use two channel paths to devices.
Make all work data sets the same size, or as nearly the same size as possible.
Make sure the SORTWKdd data sets do not share devices or channels with the SORTIN, SORTOUT, or OUTFIL data sets.
Specify contiguous space for work data sets, and make sure there is enough primary space so that the automatic secondary allocation is not needed.
Provide adequate virtual storage when work data sets are allocated on non-synchronous devices, as explained in Non-synchronous storage subsystems.

Elapsed time is decreased when DFSORT can both read input while writing to SORTWKdd and write output while reading from SORTWKdd. If, for example, you have two channels, the best allocation of them is to have SORTIN, SORTOUT, and OUTFIL data sets on one and the SORTWKdd data sets on the other.

Storage requirements for different disk techniques can be estimated by using the guidelines found in Using work space.