Hello,

First question:  Have you seen the update to the Initialization & Tuning
Guide?

https://www.ibm.com/support/knowledgecenter/en/SSLTBW_2.2.0/com.ibm.zos.v2r2.ieae100/smflim.htm

There is information there about selecting these values, and how they
relate to each other. This book was not available when we built the PDF for
OA47062. Specifically, look for a section labeled
 "Establishing SMFLIM rules in SMFLIMxx"
I will embed other responses below, tagged with SBJ:

Regards,
Steve

On 7/13/2017 2:55 AM, Vernooij, Kees - KLM , ITOPT1 wrote:
> Hi Jim,
>
> I have been digging into replacing IEFUSI by SMFLIMxx and have a few
questions:
>
> - How broad has REGIONX been implemented? It is available in JCL and
IEFUSI and SMFLIMxx can set both values, but I don't see them in SMF,
SMF30RGN still has one value, and IEFUSI WORD5/SUBWORD2 also has only one
value. I had supposed REGIONX would have been implanted in these places
too.
SBJ:  REGIONX is implemented in most places; SMF30 is the exception.  As
for IEFUSI, check out the new flags in Word 5, subword 1. There is a flag
that indicates REGIONX values are being supplied, and those values appear
as Word 5, subwords 7 and 8.>
> - What is the relation between IEFUSI's Region Limits and SMFLIMxx's
SYSRESV values. It looks as if they are each other's complement.
> Suppose my system has a userregion of 9M. if I set IEFUSI's Region below
to 6M and Regionlimit below to 7M, this leaves 2M for the system region.
SBJ: This is described better in the Init&Tune Guide, but your assessment
of the SYSRESV values is correct. You specify how much to leave for the
system to use, and we do the math to set the limit.

Note that SMFLIMxx REGIONBELOW/REGIONABOVE are *not* related to the
IEFUSI's "region size" parameter, but rather, implements a new function.

Given your example above, the maximum private area available to the program
is 7M.  But what if you have thousands of JCL jobs that specify REGION=4M
on the JOB or EXEC statements?  You can use REGIONBELOW/REGIONABOVE to
override the JCL with a value you choose, and avoid updating those
thousands of jobs. Choosing REGIONBELOW(NOLIMIT) means you'll get all of
the non-extended private storage (except for the storage reserved via
SYSRESVBELOW). REGIONABOVE works similarly.

There is no SMFLIMxx attribute that mirrors IEFUSI's (variable) region size
at this time.

> If I set SMFLIMxx's Regionbelow to 6M and SYSRESVBELOW to 1M, does this
result in a Regionlimit of 2M?
SBJ: No, the region limit will be 6M.  The SYSRESVBELOW value of 1M,
subtracted from the total non-extended size (9M) means that the total
available for the step is 8M.  Your REGIONBELOW then overrides the job or
job step's REGION value and simply set a lower cap on the limit.

> So with IEFUSI, the Systemarea is what is left from setting Regionsize
and Regionlimit, with SMFLIM the Regionlimit is what is left after setting
Regionsize and Systemarea. Correct? This is not directly clear from the
documentation.
SBJ: To restate a bit:  with IEFUSI, the system area is what is left from
setting the Regionlimit. The IEFUSI Regionsize value protects some amount
of storage from overly large variable-length getmain requests. But that
storage is still available to the user program via fixed-length getmains.
This is described in the Initialization and Tuning Guide, and was pointed
to from the IEFUSI chapter of the Installation Exits book.

> Besides that, the IEFUSI documentation explains the difference between
Regionsize and Regionlimits and their value for variable getmains.
SMFLIMxx's doc does not and one could easily decide to set values resulting
in a zero Regionlimit. Getmainproblems will then be difficult to analyze.
SBJ: Read through the Init&Tune Guide and see if that doesn't help.  I will
use the above feedback for future doc updates. >
> - In IEFUSI I can query the requested Regionsize and act on it. If the
step requests less than 7M, I give it 7M. If it requests more, I allow it
what it request. In SMFLIMxx the requested Regionsize should therefore be
available as filter together with LE/GE/EQ/NE operators.
SBJ:  I agree that a filter would be useful.  As Jim said, You should be
able to submit an RFE, or find an existing one to which to add your vote.
These are looked at by Development.>
> - As mentioned before, setting data/hyperspace limits setting is also
desired if SMFLIM is to be a replacement if IEFUSI.
SBJ: I concur. This would also make a good RFE.

>
> - Besides all this, we use IEFUSI to reset the Serviceclass of certain
jobsteps, but this is a private specialty.
SBJ:  And for this, IEFUSI would still be useful. If IEFUSI does not set
Word 5, subword 1, bit 0, none of its values will be used. And if an
SMFLIMxx rule matches the job step, it will override the exit.

>
> Sometimes we see surveys in this list if someone is planning to change
something or develop something new and is interested what functions would
be whished. If SMFLIM's intention is to replace IEFUSI, a survey of current
IEFUSI functions might have helped
>
> Thanks,
> Kees.



Regards,

Steven B. Jones
z/OS BCP Development - Allocation/Scheduler/SMF

----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN

Reply via email to