CPU and I/O Parallelism

•  Requires Partitioned Tablespaces
•  Max Parallel Processes Determined at Bind
  – Based on Partition Statistics
•  Runtime Parallel Processes Depend on Buffer Pool Resources
•  Substantial Elapsed Time Reductions
• Sysplex Query Parallelism

Again, the basic of performance have not changed. We have new facilities that may provide substantial performance improvements for large table space scans. However, the gains may be limited by CPU, Buffer Pool, and DASD issues.

I/O parallelism (CPU parallelism includes parallel I/O streams) requires that each partition be on a separate physical device or performance will not improve – in fact it might even be worse if the physical datasets are not properly spread out.