The PI limits that you specify in your provisioning criteria depend on what you would consider tolerable when capacity is constrained. If you define the goal for the service class period so that the goal can still be achieved in times of peak demand, and additional capacity does not need to be active, the provisioning PI could be set just above 1. In other cases, with more aggressive goals in effect, a provisioning PI would need to be higher.
The provisioning duration determines how fast the Provisioning Manager activates additional capacity when the PI of that service class period is consistently above the limit. The duration must be viewed relatively to the gathering interval length of the monitoring product, such as the RMF™ MINTIME, whose default is 100 seconds. In general, the duration should be longer than 3 minutes, to prevent a short-term disturbance from triggering a provisioning action. It can also take some time for WLM to resolve a PI problem by reassigning resources.
The specified duration is not a guaranteed reaction time. Certain events cause the Provisioning Manager to block some time to allow WLM to readjust. Events that can trigger a block include, for example, the activation of a new WLM service definition or policy, or a capacity change in the observed CPC.