Well, after a review of smf30 records we find that those here who mentioned IEFACTRT win the gold (even though the open cermonies don't start until later tonight in London). The IBM supplied sample member IEEACTRT for z/os 1.13 has changed the calculation for service units. The v1.11 exit 'service units' included only SRB time, while the v1.13 exit 'service units' include TCB, SRB, I/O and MSO. Here is the comment from the source that indicates there has been a change:
*/*$P7= OA31624 HBB7770 100128 PDRJ: IEEACTRT USING WRONG FIELD @P7A*/ */* FOR TOTAL SERVICE UNITS @P7A*/ */* (SUG APAR) @P7A*/ So those using the vanilla IEFACTRT should expect this new calculation. As for us, we continue chasing the other half of this issue - increased wait times for our batch jobs. We have been here before. We were cpu constrained on z/os v1.11. Two months ago we took several measures to improve our batch 'window'. And now, after upgrading to v1.13 (with no hdwr or memory changes) we have again become a bit more cpu constrained. This is my theory: z/os v1.13 = more overhead. Does anyone here claim that we should be observing *less* overhead on v1.13 compared to v1.11? Thx again for the excellent guidance. -Jim ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO IBM-MAIN
