SMS allocation services perform optimally when you define a limited number of storage groups to contain your volumes. However, you might want to isolate large data sets in their own storage group because of their unique space management requirements. Additionally, by separating large data sets from other data sets, you prevent allocation failures that can occur due to volume fragmentation, since allocation can fail if the requested quantity cannot be satisfied in five extents.

The space requirements for large data sets can limit the free space available to other data sets. Also, more space must be reserved on volumes to support the new allocation or DFSMShsm recall of large data sets. Because of this, the high allocation/migration thresholds for storage groups containing large data sets should be set lower than for storage groups containing normal-sized data sets.

Table 1 provides a list of recommended sizes to determine what constitutes a large data set, according to the type of DASD volume.

Your own definition of large should provide for successful DFSMShsm recall in five or fewer extents, based on the threshold you have set for the storage group.

Management class can be used to limit the negative effects of large data sets during the space management process. As an example, you can migrate large data sets directly to migration level 2, rather than letting them move to migration level 1.

Another management option is to place large data sets in your primary storage group and assign a management class that prevents migration. This is a good solution if you have only a few large data sets that need to be immediately available on DASD.