cpu% RSCL vs. SCL (RPC), Enclave cpu

Resolving The Problem

The reason for the difference is due to the CPU% in RSCL being calculated from consumed and available CPU service units for the period while the CPU% for SCL is based on the rolled up CPU time consumed by the individual address spaces in the report/service class for the period . Enclave cpu may be a major factor for the divergence of these two values.

The rolled SCL cpu time includes SRB, TCB and ASSBPHTM (enclave preemptable SRB time- cpu time for preemptable SRBs (enclaves)) that were executing within that asid. The SRB may be executing on behalf of another address space which is in a different service class period.

RSCL cpu is determined from a particular period's CPU service units from the smf type 72 records. Enclave cpu is accumulated in the service class period the enclave is assigned to. As explained in 'Enclave Resource Accounting' in 'z/OS V1R4.0 MVS Programming Workload Management Services', an enclave may originate in one service class period and be assigned to a different service class period. If the enclave is not assigned to a different service class period then the owning period is the assigned period for the enclave. The asid the enclave executes in may not be in the assigned service class period for the enclave.

So e.g. this is what you can have:

In the above example SCL will show cpu time for enclave1 (a dependent enclave) in service class period 2 and 3 . RSCL will show the cpu time for the enclave 1 in service class period 1. For enclave 2 SCL and RSCL will account for the enclave cpu time in service class period 2. For enclave 3 SCL and RSCL will account for the enclave cpu in service class period 1. (Note that if you were looking at SMF type30 records the enclave cpu or all 3 enclaves would be accounted by the owning asid in serviceclass period 1)

SCL time is being captured in EXDC while the cpu service units for RSCL are captured by the Epilog collector from the SMF 72.3 record If the CMS or Epilog collector STC is not up or connected for the reporting time in question that can cause discrepancies.

Note that on panel KM2CP1D (Omegamon II CUA) the enclave cpu% shown for an address space is cpu for enclaves currently executing within the address space shown (ASSBPHTM time), which may not be the owning address space for the enclave. As for all cpu% on this panel the enclave cpu% represents the cpu used during an interval (refresh to refresh).