RMF Introduction to LPAR Soft-Capping
Workload charging enables data centers to use both hardware and software to manage costs and system capacity: customers determine the portion of processing capacity needed to run their workloads; and, for the selected products used to run these workloads, charges are based on a site-defined capacity, rather than on the entire CPC capacity.
rm
The measurement unit used to define and monitor an LPAR capacity is the MSU (millions of service units per hour), which is calculated as follows:
MSU = ( su/sec * Nb_of_CPs * 3600 ) / 1000000
It is important to understand that, depending on the way you size your logical partitions, both software charges and system performance will be impacted. When under WLM control, the system will continuously monitor the actual MSU consumption of an LPAR, and, depending on the LPAR size, will restrict the partition from using more CPU resource than its allowed capacity.
The method of assigning a defined capacity to an LPAR, also called a soft cap, overcomes the traditional capping methods limitations. It also allows spikes in your workload: when an LPAR has a defined capacity, the Workload Manager (WLM) compares this limit to the average MSUs consumed during a four-hour rolling interval. As long as the four-hour rolling average stays below the defined capacity limit for the LPAR, utilization is allowed to spike. As soon as this average rises above the defined capacity limit, WLM restricts the partition from using more CPU resource than its defined capacity limit, and the LPAR remains capped (restricted) until the four-hour rolling average drops below the defined capacity limit.
This is the recommended method for setting up Variable Workload License Charges (VWLC) for a product: charges are based on the defined capacity; spikes in your workload, up to the total LPAR size, are allowed. However, correctly sizing your LPARs capacity limits (and hence, your ability to achieve savings on software license charges by tailoring your license certificates) requires the tracking of your peak four-hour rolling average values and the WLM capping percentage.
If the peak value is constantly lower than the defined limit, you may consider decreasing the LPAR capacity until WLM begins capping the partition (a capping percentage of 1% to 5% ensures the work is done optimally at the lowest cost). Capping affects all workloads running in the partition and can result in elongated response times and generally unacceptable performance.
If the peak value exceeds the defined capacity for long periods (high WLM capping percentage reported), you may consider increasing the LPAR capacity limit until the WLM capping percentage falls to within an acceptable range.
If the peak value occasionally exceeds the defined capacity (low WLM capping percentage), you may consider shifting some workloads to a time of day when the LPAR is less busy.
Ideally, your peak value should always be as close as possible to the defined capacity, without exceeding it, while your WLM capping percentage should always be near to zero.