This workspace provides HiperDispatch-related topology and utilization statistics at the LPAR and individual logical processor level for the LPAR on which an OMEGAMON XE on z/OS agent is running when the LPAR is in HiperDispatch mode. This information can help you determine how your HiperDispatch configuration is performing.
HiperDispatch addresses make efficient use of CPU-level hardware caching in processors that employ Non-Uniform Memory Access (NUMA) in their designs. NUMA is a means of minimizing memory access latency when retrieving CPU-level cache contents. Efficient use of NUMA architecture requires the dispatching of work, where possible, on the same physical CPU, or another CPU on the same book, to take advantage of cache hits in L1, L1.5 or local L2 caches. Collaboration between the hardware, the LPAR Hypervisor and the Workload Manager Dispatcher is necessary to fully take advantage of what NUMA offers regarding workload performance.
HiperDispatch Management at the z/OS image level can be deployed on one or more of the LPARs in a complex. HiperDispatch is activated for an LPAR by specifying HiperDispatch=YES in the SYS1.PARMLIB IEAOPT member. HiperDispatch=NO, the default, physical CPUs will be assigned equally across all of the logical CPUs in an LPAR.
For the HiperDispatch feature to be used, the following conditions must be in effect:
WLM can then determine if the LPAR should be in HiperDispatch mode or not. WLM will place the LPAR in HiperDispatch mode if there is sufficient work in this LPAR to keep 1.5 physical CPUs busy. If demand falls below this threshold WLM will take the LPAR out of HiperDispatch mode.
Two views present information at the LPAR level:
When HiperDispatch Management is on, three additional views display the logical CPU ID, entitlement (High, Medium or Low), percentage entitlement, physical CPU percentage, physical overhead percentage managing the CPU within the LPAR, and a status (Online, Offline, Parked, Park Pending, or Reserved) for each CPU, by logical type.
You can link to this workspace from a row in the System CPU Utilization workspace.
See also: System CPU Utilization workspace, Organization of system-level predefined workspaces, Attribute groups used by the system-level predefined workspaces