Moving data in parallel

The author examines a method of applying parallelism to data-moving operations to enhance performance so that they may fit into today's maintenance windows. She specifically discusses converting the algorithm used to load an alternate key file (index) from serial to parallel using Tandem's Non-Stop SQL. It is shown that using disks as the unit of parallelism allows data movement to be parallelized even when ordering is necessary. The techniques used to translate the index loading algorithm from serial to parallel can be applied to other cases in which records are moved. So long as machine cycles are available this approach to parallelism can provide significant improvement.<
>