Migration

Retaining certain types of data on ML1 DASD or ML2 tape is undesirable. Migrating those data sets to the cloud can be desirable when the data sets are large, and not expected to be referenced for a long time.

When you migrate a data set to the cloud, it uses offboard data movement that is provided by the hardware. The elapsed time for data set migrations and recalls to and from cloud storage are longer than the time for migrations and recall functions on FICON attached storage. With cloud storage, data transfer is performed with TCP/IP across Ethernet connections. IBM recommends that installations perform tests to determine the expected elapsed time degradation unique for their environment. You can use DFSMShsm FSR records to determine both elapsed time and CPU consumption for these functions. Based on the expected degradation, migrate to cloud storage only those data sets for which the degradation is acceptable. For example, migrated data sets that must be recalled within a specific time frame to achieve an application SLA, should be migrated to cloud storage only when the expected degradation is acceptable. Data set types that are both large in size, rarely or never recalled, are recommended for migration.

Because the backup format is different between an ML1/ML2 migration copy and a cloud migration copy, a data set cannot be moved from ML1/ML2 directly to the cloud. First, recall the data set to L0 DASD, then remigrate the data set to the cloud.
Note: When the catalog entry for the data set is migrated to the cloud, it has a volser of MIGRAT, like other migrations. However, the device type field in the catalog entry is x'00018000'. The third byte x'80' indicates that it is a tape, and the second byte x'01' indicates it is a cloud type.