The concept of HiperDispatch mode
Without HiperDispatch, for all levels of z/OS®, a TCB or SRB may be dispatched on any logical processor of the type required (standard, zAAP or zIIP). A unit of work starts on one logical processor and subsequently may be dispatched on any other logical processor. The logical processors for one LPAR image will receive an equal share for equal access to the physical processors under PR/SM LPAR control. For example, if the weight of a logical partition with four logical processors results in a share of two physical processors, or 200%, the LPAR hypervisor will manage each of the four logical processors with a 50% share of a physical processor. All logical processors will be used if there is work available, and they typically have similar processing utilizations.
With HiperDispatch mode, work can be managed across fewer logical processors. A concept of maintaining a working set of processors required to handle the workload is introduced. In the previous example of a logical partition with a 200% processor share and four logical processors, two logical processors are sufficient to obtain the two physical processors worth of capacity specified by the weight; the other two logical processors allow the partition to access capacity available from other partitions with insufficient workload to consume their share. z/OS limits the number of active logical processors to the number needed based on partition weight settings, workload demand and available capacity. z/OS also takes into account the processor topology when dispatching work, and it works with enhanced PR/SM microcode to build a strong affinity between logical processors and physical processors in the processor configuration.