Peter Hunkeler wrote: >I'm analysing a perfomance issue with long running (35 hours) a batch job. >Looking at JESYSMSG, I stumbled across information that our IEFACTRT exit >writes at step end. There is one step that seamed to do millions of I/Os >against STEPLIB. STEPLIB is a two DSN concatenation, however, I see dozens of >STEPLIB related I/O numbers....
>Trying to understand, I started to have a detailed look at the SMF 30 records >for the step. We do have interval recording, so I can see one subtype 1, a >bunch of subtype 2, followed by one subtype 3 for the lat interval, and >finally 1 subtype 4 records at the end of the step. Do you also see subtype 5 (Job termination)? >Q1: The I/O counts in the interval records are *not* cumultive, but reflect >the count since the previous record has been cut, correct? This is how I read >the doc. This is also how I read it. AFAIK, step 5 is the cumulative one - counting all steps, but you want details PER step. >Q4: If Q1-Q3 are correct, this explains why IEFACTRT listing I/Os by DDname >and device can shoe dozens or hundreds of EXCP numbers for a two DSN STEPLIB. What is your DDCONS and INTERVAL setting? Groete / Greetings Elardus Engelbrecht ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO IBM-MAIN
