The CPU time consumed in ACS is recorded in the SMF 30 step termination
record in the "Initiator" or "Privileged" TCB and SRB fields SMF30ICU
and SMF30CPS (MXG variables CPITCBTM and CPISRBTM), although it's not
just ACS CPU time that is recorded in those fields:
Those fields contain all captured CPU time from step initiate until
the problem program was loaded and began execution, plus the CPU
time from program end until step terminate.
The "front end" CPU time includes JES2 DSENQ (first step only),
catalog search, ACS execution, and actual allocation, which could
include set up costs for HSM recalls, and probably other things.
The "back end" CPU time is the CPU time to build the final subtype
3 and subtype 4 type 30 records, and that can be significant if
there are thousands or tens of thousands of DD statement in the
type 30 (long-running STCs that dynamically allocate, SYSOUT
processors, etc.). But it will be horrific if you still have the
IBM default DDCONS=YES in SMFPRMxx in SYS1.PARMLIB.
You should always specify DDCONS=NO, and to prevent your
long-running STCs from running out of private area, you
must always specify NODETAIL for your STCs.
You should be able to compare the CPU costs of ACS rule execution
with a series of steps (start with step 2 to eliminate the cost
of DSENQ in JES2) with your JCL and PGM=IEFBR14.
Barry Merrill
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO
Search the archives at http://bama.ua.edu/archives/ibm-main.html
