Backup performance (CFINT)

[bvining@xxxxxxxxxxxx]

In my opinion...

I would have to say that the performance people are probably closest
to the mark.  If the interactive threshold is hit, CFINT should
kick in.  Note that, as pointed out in other notes, CFINT can also
"kick in" for a variety of other reasons and not just because the
interactive threshold has been reached (for instance CFINT is used
as a general purpose catch-all for some system activities that cannot
be readily associated with a specific task even though CFINT itself is
not actually executing).

Having said the above, a definite gray area can exist in that if one
(and I do mean one) interactive task is active and the interactive task
has minimal interactive interaction, then the interactive threshold is
unlikely to be reached and so CFINT should not kick in.  The "gray area"
has to do with defining "minimal interactive interaction" and is a bit
too much to really get into.  This area of one interactive task
executing is independent of the system being in restricted state (though
making sure just the one job is active is obviously alot easier to
control by going into restricted state), so having other selected
subsystems up concurrent to one save operation should be OK.  You could
be in restricted state and still see CFINT per the above (if the job
is heavily interactive, though then I wonder if a traditional application
could really drive the CPU if waiting for terminal input and performing
normal IO operations).

Bruce

>
>"Al Barsa, Jr." wrote:
>
>> This is an interesting subject.  The performance people that I know at IBM
>> claim that CFINT hits whenever you exceed the interactive threshold.  The
>> system managers (who claim to know more about performance than the
>> performance people [if you believe that]) disagree, and say that CFINT
>> doesn't happen in a restricted state.
>>
>> Bruce Vining, do you have an opinion?
>
>I don't think it came thru in my original post very well, but I don't
>have a totally restricted system during backup either.  I end QINTER,
>QBATCH, and QSERVER subsystems, as well as all TCP/IP and host servers.
>Subsystems QCTL, QCMN, QSPL, QSYSWRK, QSVCDRCTR, etc. stay up.
>
>-Jeff
>


+---
| This is the Midrange System Mailing List!
| To submit a new message, send your mail to MIDRANGE-L@midrange.com.
| To subscribe to this list send email to MIDRANGE-L-SUB@midrange.com.
| To unsubscribe from this list send email to MIDRANGE-L-UNSUB@midrange.com.
| Questions should be directed to the list owner/operator: david@midrange.com
+---