RMFOD4 - Daily z/OS LPAR MSU Use Cap and 4 Hour Avg
The RMFOD4 inquiry shows the average MSU usage of z/OS LPARs at the RMF interval level and includes the four hour rolling MSU average.
micsrm140
All z/OS LPARs are accounted for. This inquiry is similar to the RMFOD5 inquiry but has the additional benefit of showing the four hour rolling MSU average which is only available for z/OS LPARs. The inquiry generates two separate charts. One for z/OS LPARs that have defined capacity limits, and the other for z/OS LPARs with no defined capacity limits. One of the uncapped LPARs for each CPC is the ‘PHYSICAL’ LPAR that captures PR/SM overhead.
The top example shows the average MSU usage chart for a z/OS LPAR (PRSMLPNM=’CA31’) running on an IBM zEC12 2827-615 CPC. The RMF intervals records were generated at 30-minute intervals, and the LPAR has a defined capacity limit of 300 MSUs. The second example shows the average MSU usage chart for another LPAR (PRSMLPNM=’ZQ01’) on the same CPC. The RMF intervals records were also generated at 30-minute intervals. This second LPAR has no defined capacity limit.


Uses
The processing capacity of mainframe CPCs is expressed in MSUs (Millions of Service Units per Hour). One of the characteristics of a specific CPC model is its MSU capacity. This capacity is indicative of its workload processing capability and is directly related to the number of general-purpose CP engines configured. Each LPAR that is assigned logical CP processors uses some amount of the available CPC MSUs. The MSU capacity of a CPC and of individual LPARs can influence the cost of software licensed to run on LPARs.
Some software licensing costs can be lowered by defining a capacity limit for an LPAR. The limit prevents the LPAR from using the full MSU capacity associated with the logical processors assigned to it. The RMFOD4 charts can be used to monitor the normal MSU consumption of LPARs that use general-purpose CP engines, and also to determine the appropriate value to use for a defined capacity limit. The four hour rolling average is helpful because LPARs are allowed to exceed the defined limit, without penalty, as long as the four hour rolling MSU average stays below the defined limit.
For example, if the daily MSU usage shown for the LPAR in the top example (PRSMLPNM=’CA31’) is consistent, over time, it means that the limit of 300 MSUs is set too high. A defined capacity limit of about 280 MSUs is a more appropriate setting for this LPAR - given that the peak four-hour rolling average was 276 MSUs.
Charts
Daily MSU Use for Cap z/OS LPARs 4H Avg
All LPARs in this data extract have defined capacity limits. If no LPARs in any CPCs have defined capacity limits, no charts are produced.

Left Y-axis
CPUAVMSU
Avg Number of Consumed MSUs
Right Y-axis #1
CPUAVLAC
Long-Term Average of CPU Service (MSUs)
Right Y-axis #2
CPUDEFLM
Partition Defined Capacity Limit
X-axis
HHMM
Hour : Minute
Data element HHMM is derived by concatenating the HOUR and MINUTE of each RMF interval together, separated by a colon, HH:MM (for example, 14:59 meaning hour 14, 59th minute).
Daily MSU Use for Uncap z/OS LPARs 4H Avg
All LPARs in this data extract have no defined capacity limits. There is always at least one LPAR in this data extract for each CPC -- the ‘PHYSICAL’ LPAR which shows PR/SM overhead.

Left Y-axis
CPUAVMSU
Avg Number of Consumed MSUs
Right Y-axis
CPUAVLAC
Long-Term Average of CPU Service (MSUs)
X-axis
HHMM
Hour : Minute
Data element HHMM is derived by concatenating the HOUR and MINUTE of each RMF interval together, separated by a colon, HH:MM (for example, 14:59 meaning hour 14, 59th minute).
Inquiry Defaults
- HARCPU DETAIL timespan file: Cycles 01 - 02
- Filters: PRSMUSCT GT 0 (only PR/SM managed LPARs)
- Previous Day is selected
- Derivations:
CPCID
%CPCID;
DATE
DATEPART(ENDTS);
HHMM
PUT(HOUR(ENDTS),Z2.) || ':' || PUT(MINUTE(ENDTS),Z2.);
Modifications
Modification considerations:
- If you want to see hourly, rather than interval level MSU usage and four-hour rolling MSU averages, you can change use the HARLPC DAYS timespan.
- You can increase the number of cycles to create charts for multiple days, but remove the step that selects only yesterday’s data.
- Using the following CPC Identification data elements you can filter on CPC:
- fffMOD - Processor Model Family (where fff is the file identifier -- for example., CPU, LPC, and so on.)
- CPCMODID - CPC Model Identifier
- CPCSEQNB - CPC Sequence Number