Working with a table space that is already partitioned
You can examine an existing partitioned table space to determine whether parallel processing can be improved.
About this task
Assume that a table space already has 10 partitions and a particular
query uses CP parallelism on a 10-way CPC. When you add other read I/O
wait time
(from accounting class 3) and processing time (from accounting
class 2), you determine that I/O cost is three times more than the processing
cost. In this case, the optimal number of partitions is 30 (three times more
I/O paths). However, if you can run on a data sharing group and you add another Db2 subsystem to the group
that is running on a 10-way CPC, the I/O configuration that minimizes the
elapsed time and allows both CPCs to run at 100% would be 60 partitions.