Funny you should mention this....
I recently converted our daily reporting from type7x to type30 SMF records.
We are z/OS 1.8, our iefactrt doesn't do anything to the smf records (it just
writes out a message that contains the return code - I should know, I wrote
that exit, from the quite recent 1.8 samplib).
One requirement was to list the top 25 cpu using batch jobs per system (from
the smf30-5 records). We looked at those in the joblog, and we saw the exact
same behaviour: TCB time in one step was *a lot* bigger than elapsed time for
the complete job! Also cpu time consumed from the *complete* job was less
than what was shown for this one step as tcb time.
No we have neither ziips nor zaaps, but these lpars all have more than one cp.
I had explained it away as a multitasking application ("we need to go and ask
them how they use their tcbs, which they probably don't even know"), and
unfortunately I cannot find that job again on a cursory search.
Are you all saying that this might be a bug and I should dig again? :-(
Regards, Barbara Nitz
----------------------------------------------------------------------
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
