If the capacity share exactly equals the defined capacity limit, WLM instructs PR/SM to fully cap the partition at its weight.
If the capacity share is less than the defined capacity limit it is not possible to permanently cap the partition at its weight because the partition would not be able to use the capacity it is entitled to. WLM defines a cap pattern which caps the partition at its weight for some amount of time over an interval and then removes the cap during the remaining time of the interval. On average this appears as if the partition was constantly capped at its defined capacity limit. The cap pattern depends on the ratio of the capacity share to the defined capacity limit. If the capacity share based on the weight is rather small compared to the defined capacity limit the partition will be capped very drastically during short periods of time. Therefore, for configurations requiring the capacity share to be smaller than the defined capacity limit, it is recommended to keep both definitions as close as possible.
If the capacity based on the weight is greater than the defined capacity limit, WLM instructs PR/SM hypervisor to define a phantom weight, because it is not possible for WLM to cap the partition at its capacity based on the weight. A phantom weight pretends a utilization of the CPC for a particular partition which makes it possible to cap the partition to be managed at the defined capacity limit.
Options 1 and 3 are the recommended ways of specifying defined capacity limits and weights because they provide a capping behavior which is very smooth.
zEC12 (GA2) and later systems support the use of negative phantom weights in Option 3. For systems running on eligible hardware with the required software support (z/OS® V2R1) option 2 above is obsolete and option 3 will be used. Consequently the smoother flavor of capping will always be used regardless of the ration of the capacity share to the defined capacity limit.