Performance improvements for the LOAD utility
In Db2 11, the elapsed time of some LOAD utility jobs is reduced because LOAD can use parallelism in more situations than in earlier versions of Db2. LOAD can use parallelism for partitioned table spaces with only one input data set. Also, LOAD can use parallelism for non-partitioned table spaces.
In earlier versions of Db2, LOAD uses parallelism for partitioned table spaces only if you provide a separate input data set for each partition. Providing and maintaining such data sets is difficult, especially if you altered the partition boundaries and then had to redistribute the data across the input data sets. Also, LOAD cannot use parallelism for simple or segmented table spaces.
In Db2 11, these restrictions are removed. You no longer need to provide separate input data sets for each partition so that LOAD can use parallelism. Also, LOAD can use parallelism for non-partitioned table spaces. Specifically, LOAD can use parallelism with one input data set for all of the following table space types:
- Simple table spaces
- Segmented (non-UTS) table spaces
- Partitioned (non-UTS) table spaces
- partition-by-range table spaces
As a result, the elapsed time for the load operation can be reduced.
To specify that you want LOAD to use parallelism if possible, use the new PARALLEL keyword in the LOAD utility control statement.