High speed online copy of partitioned data

Abstract

Disclosed is a CPU-implemented method for logically copying partitioned datasets from a source device to a target device in a storage subsystem. The method improves performance over prior logical copy utilities by substantially decreasing the number of I/O's required for the operation. A source dataset is designated on a first storage device and members are selected for copying. A target dataset is designated on a second storage device. The target dataset may be newly created or may already exist. If the latter, existing members remain alterable by other system users during copying. All selected source members, or a subset thereof, are opened for input. An equal number of target members are created in the target dataset in a single directory write operation. All data records of the selected source members are input to the CPU's virtual storage, and then transferred to the target members via frugal use of data I/O's. When the selected data records have been transferred, all target members are named using a single directory write operation. Finally, the source and target members are closed. During copying, other system users may browse the selected source members, but the unnamed target members will be invisible to them until copying is complete.