Yes, you are right. The Maximum Pool Capacity defines an upper limit up to which the underlying partitions can consume the processor capacity. The micro-partitions in a shared-processor pool can’t get additional capacity even if there is spare capacity available in other shared-processor pools if its maximum capacity is exhausted (Page 16).

In your example, since the maximum poor capacity is 3, both partitions in total can get maximum 3.0 processing units (and not 3.6) ,subject to the availability of the processing capacity.

Just to elaborate further that why we need mutiple shared-processor pools at the first place. The reason is – because you want to prioritize a set of partitions to get more processing capacity without compromising with the number of active partitions (you can prioritize using entitled capacity but it would restrict number of partitions you can start at a time) and at the same time have the flexibity to support the spikes in utilization (using uncapped partitions and high value of their shared-processor pool’s maximum pool capacity). In the absence of multiple-shared processor pool you have to leverage entitled capacity and uncapped weight but that way all the uncapped partitions would compete for free CPU cycles and not just your set of “favourite” partitions. In multiple-shared processor pool you define a set of partitions and want to first distribute the available cycles first among them before donating it to public partitions.

It is important to note that default shared-processor pool’s maximum pool capacity is always equals to active physical processors.

I hope it helps!

Cheers, Neeraj

I read your article “Capacity Planning and Performance Management on IBM PowerVM Virtualized Environment” since I’m planning to implement shared pools on a 770. One thing is unclear to me about the Max Pool Capacity. Is it so that the sum of all capacity for the LPARs in this pool whether obtained through Level-0 and/or Level-1 can never exceed the Max Pool CPUs?

Example: I have a pool with 2 uncapped LPARs, per lpar 0.5 Entitled CPU, 2 Virtual CPU. The pool has 3 CPUs max. Assumed that both LPARs at some point want to consume 1.8 CPU each, can these LPARs then max get 3 CPU and not the requested 3.6?

Many thanks,
Rob