Ron and Jenny Hawkins wrote:
Tom, et al,
I was just thinking... CICS aside, when a job ABENDS while updating a KSDS
file, what is the most common thing that happens to allow the job to be
rerun? Delete and restore the KSDS, right? So what good are the accumulated
statistics at this point - valid or invalid? Well there worth zilcho,
because they were just deleted. Gone forever.
The ABENDing job steps can also use VSAM datasets READ only, in which
case the stats will be inaccurate but no restore is needed. Even in
case of an updated file, it can depend on the application and nature of
updates whether a restore is needed or not.
And the next day someone takes a LISTCAT of that file and makes some
decision based on those stats, assuming that it represents 24 hours of
activity, when it is really just the last 20 minutes of the batch run. Smart
move!
If those 20 minutes of stats show enough CI splits to justify a
reorganize of the file, it doesn't matter that this only represents 20
minutes of activity. If the CI-split count is above your reorg
threshold, it also doesn't matter if you lost 15 hours of CI-split
counts because of a CICS crash, a reorg is still justified. You may be
late with a decision to reorganize, but better late than never. With no
stats at all, an automated process has no basis for choosing whether to
reorganize or not, and arbitrarily choosing to reorganize could be a bad
choice for a large, rarely-updated VSAM file - especially if doing an
unnecessary reorganize delays the availability of a critical online
application.
I would rather we just got rid of all the stats from LISTCAT. We don't have
a listdb2, listdl1, listfastpath or listqsam, so why the religious fervour
around LISTC stats?
Ron
LISTCAT stats have been around for at least several decades, maybe since
the beginnings of OS/360. Automated maintenance procedures and
heuristic guidelines have been developed based on LISTCAT statistics and
we have become dependent on them, because you build tools based on what
is available, even if not perfect. Without LISTCAT stats, we would have
to expend resources to rethink and redesign a number of automated batch
processes that have been working reasonable well up to now (z/OS 1.4).
...
--
Joel C. Ewing, Fort Smith, AR [EMAIL PROTECTED]
----------------------------------------------------------------------
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
