"The part of file allocation that has significant cost is basically the same whether the allocation is via JCL or dynamic allocation."
While the true cost in CPU seconds for a dynamic allocation might be exactly the same as for a static allocation, where those CPU seconds are recorded in RMF and SMF are quite different. The "Initiator" CPU time records the CPU time for allocation of all of a step's static DD statements (CPITCBTM/CPISRBTM/SMF30ICU/SMF30ISB). This is the CPU time from initiate until PROGRAM FETCH loads the program at LOADTIME, and includes the cost of the ACS allocation rules, static allocations, static HSM recalls, and possibly VAM and SMS Trace. But these "Initiator" CPU time also records the CPU time for creation of and writing out of the SMF records, which occurs after program terminate and before SMFTIME. -The "Initiator" CPU time is ONLY recorded in SMF 30 records, and only in the Step Terminate records. -They are NOT recorded in SMF 30 Interval records. -They are ONLY recorded in SMF 30 Termination records. -THEY ARE NOT RECORDED IN RMF TYPE72GO Service/Reporting Class records. -They are NOT shown in any IBM messages printed on the Job log. Instead, the CPU time for dynamic allocations occur after the step's program has been loaded, so that CPU TCB/SRB time will be recorded in the "normal" CPUTCBTM/CPUSRBTM/SMF30CPT/SMF30CPS CPU time variables, which ARE recorded in SMF 30 interval and termination records, and in RMF 72 service classes. So comparing dynamic to static allocation requires looking at all seven CPU times in the two step's type 30 termination records. Note that prior to z/OS 1.12 there was only one TCB and one SRB bucket for the "Initiator CPU time", which contained both the "INIT" and "TERM" CPU times. In z/OS 1.12, the SMF 30's now contain four new CPI buckets, a TCB/SRB pair for the 'INIT' time, and a pair for the 'TERM" time, so you can get much closer to the cause of high CPI times. This will make it possible to assign these CPI CPU times into the correct interval when they were consumed, and thus can be used to improve capture ratio metrics. But: CPI CPU times are NOT captured in RMF type 72 subtype 3 records, so if you use the CAPTURAT in RMFINTRV (or compare 70 to 72 Service Class), it does NOT include any of the CPITCBTM/CPISRBTM, which must be considered if you are concerned with low RMF capture ratio. Barry Merrill ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@bama.ua.edu with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
