RMF Meaning of "CPU Busy" and "CPU Wait"

From the point of view of z/OS running in a logical partition, CPU busy time is the sum of the durations of the periods during which the logical processor is actually executing instructions. For shared processor wait-disabled partitions, this is equal to the PR/SM dispatch time. For wait-enabled partitions, however, CPU busy time is equal to the dispatch time minus the CPU wait time recorded in the CPU Data section.
rm
From the point of view of z/OS running in a logical partition, CPU busy time is the sum of the durations of the periods during which the logical processor is actually executing instructions. For shared processor wait-disabled partitions, this is equal to the PR/SM dispatch time. For wait-enabled partitions, however, CPU busy time is equal to the dispatch time minus the CPU wait time recorded in the CPU Data section.
From the point of view of the entire complex, on the other hand, central processor utilization does not depend on the wait assist status. If a logical processor is dispatched, the central processor it uses is unavailable for other use, whether it is actually executing instructions or is in the wait state. The amount of time during which the central processor is unavailable to other partitions or for LPAR management functions is equal to the dispatch time in both cases. In order to differentiate this measurement from "CPU busy" as described previously, we can think of the dispatch time as "CPU allocation" time.
CPU wait time in the LPAR environment can be defined as "CPU not busy". This covers the amount of time represented by the difference between the overall duration of the measurement interval and CPU busy time. For wait-disabled partitions, this is the duration of the measurement interval minus the dispatch time. For wait-enabled partitions, this is the interval duration minus the difference between the dispatch time and the CPU wait time recorded in the CPU Data section.
The following formulas can be used to calculate processor utilization statistics:
  • For partitions with dedicated processors:
                  INTERVAL - WAITTIME    PCT_BUSY = ------------------- x 100                    INTERVAL      PCT_ALLC = 100 (constant)                 WAITTIME    PCT_WAIT = -------- x 100               INTERVAL      PCT_NBSY = PCT_WAIT
  • For partitions with shared processors and with wait-assist disabled:
                  DISPTIME    PCT_BUSY = -------- x 100               INTERVAL      PCT_ALLC = PCT_BUSY      PCT_WAIT = 0 (constant)                 INTERVAL - DISPTIME    PCT_NBSY = ------------------- x 100                    INTERVAL
  • For partitions with shared processors and with wait-assist enabled:
                  DISPTIME - WAITTIME    PCT_BUSY = ------------------- x 100                    INTERVAL                 DISPTIME    PCT_ALLC = -------- x 100               INTERVAL                  WAITTIME    PCT_WAIT = -------- x 100               INTERVAL                 INTERVAL - (DISPTIME - WAITTIME)    PCT_NBSY = -------------------------------- x 100                           INTERVAL
    where:
                                           CA MICS label      meaning                     element ---------  --------------------------- -------- PCT_BUSY   Percent processor busy      IPUPCBS PCT_ALLC   Percent processor allocated IPUPCDT PCT_WAIT   Percent processor wait      IPUPCVW PCT_NBSY   Percent not using processor IPUPCWT INTERVAL   Measurement interval        DURATION WAITTIME   processor wait time         IPUWTM DISPTIME   processor dispatch time     IPUDTM