Sorry for the late reply; been busy thinking about all your good advice and reading the papers mentioned. I do have a much better understanding now, I think.
BTW, I also found a document from Don Deese called "Introduction to HiperDispatch Management Mode with z10" in my document archive (http://www.cpexpert.com/HiperDispatch%20Management%20Mode.pdf). Very interesting reading as well. One question that has not been answered: Are tasks from a single address space, i.e. subtasks from a multitasking program, assigned to a single "dispatcher affinity node" (or whatever this is called), or can some subtasks be assigned to one node while others are assigned to another node? I undestand that work units can be moved to other affinity queues if z/OS decides this will help. But is this only for work units from different address spaces? Just curious. To summarize the difference with or without HiperDispatch: - With HiperDispatch=NO (or before HiperDispatch) a certain workload could use all available capacity of all the CPs assigned to the LPAR, even if less capacity was sufficient to achieve its WLM goal. - With HiperDispatch=YES, a certain workload will only be able to use all the available capacity of the CPs assigned *to the affinity node*, as long as it is able to achieve or even overachieve its WLM goal with this capacity. This is true even if there was spare capacity in other affinity nodes. WLM will not unpark low processors nor reassign work to different nodes, nor will processors from other nodes help out, as long as the goal is achieved. -- Peter Hunkeler ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
