On Mon, 18 Aug 2008, Jan Vanbrabant wrote: > *** Cross-posted in the IBM-MAIN & IMS-L listservers) *** > > Hi, > > Either Iâm getting old (or at least too old to understand), or Iâm >stupid, or Iâm too tired to understand (after a long day gazing at my laptop screen) ⦠> > One way of doing accounting for billing purposes in IMS is based on the > DLRTIME field in the Application termination accounting log record > (xâ07â). (Thereâs also DLREXTIM field, which is new in IMS V10, > while we are V9. Anyway this is not important in this discussion.) IMS > TYPE07 log record CPU timer-unit field, DLRTIME, indeed allows to > capture the CPU time used by a running transaction. This execution time > is expressed in so-called âTimer Units (TU)â. TU is a z/OS notion > (STIMER-, STIMERM-, TIME-, TTIMER-macros), not an IMS concept. > > ASKQQA item RTA000060992 (19920318) about âTIMER UNITS (TU) conversion > to CPU timeâ says: <One timer unit (TU) is approximately 26.04166 > microseconds. You can multiply that value with the number of TUs > reported in the type x'07' log record.> > > Looking into the latest SA22-7607-13 zOS V1R9.0 MVS Programming > Assembler Services Reference Vol 2 (IARR2V-XCTLX) (iea2a981) (April > 2008) timer units still are âapproximately 26.04166 microseconds per > unitâ ⦠After more than 16 years ago with a couple of technology > changes & much faster processor cycles, a TU still equals 26.04166 > microseconds??? There is something I donât understand â¦. > > The real question is: how can I take into account the technology > evolution from a pre-z to a z machine (for example) in the > accounting/billing scheme? Pls shed some light in the darkness of my > mind. > > Jan >
A timer unit is independant of the resolution of the hardware clock or speed of the processor. A timer unit on a z10EC is the same value as a timer unit on a 9672. IBM designed it that way. Just as the TOD clock on all s/370s and above ran at the same speed. Bit 51 was/is guaranteed to increment by 1 every microsecond. The bits 52-103 may or may not be updated at a higher frequency. But regardless, bit 51 is incremented by 1 every microsecond. True since the TOD clock was first introduced. -- Q: What do theoretical physicists drink beer from? A: An EIN stein. Maranatha! John McKown ---------------------------------------------------------------------- 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
