COPYRIGHT (C) 1984-2019 MERRILL CONSULTANTS DALLAS TEXAS USA
MXG NEWSLETTER ONE
CONSOLIDATION OF IMPORTANT INFORMATION IN MXG NEWSLETTERS 1 THRU 8
This consolidation for new MXG supported customers will be replaced by
the MXG Supplement, to be printed in early 1987. Only the portions of
newsletters ONE thru EIGHT which are still relevant are included herein.
Newsletter Nine was either received with Version 4.4 when MXG was
purchased from SAS Institute, or is included in the separate package
from Merrill Consultants with the (empty) 3-ring binder for the
Supplement. Complete version-by-version changes are in the MXG
software, members CHANGEnn, and in members DOC.... BWM:3Feb87.
MXG NEWSLETTER VOLUME I NUMBER THREE OCT 1, 1984
2. Technical Information
a. 3800-3 Printer: A number of new variables were added to data set
TYPE6 to support the 3800-3 printer. Most welcome is DOCLENFT, the
length (in feet) of the document printed. This value has been requested
since 3800's were first announced, and it should permit accurate
cost recovery of 3800 paper. The form number was also expanded to eight
bytes (in TYPE6 FORM, and in TYPE26 OUTFORM). The 3800-3 is
supported in Version ONE of MXG.
b. CICS 1.6.1: Changes in transaction data (CICSTRAN) now provide
separate counts of message characters input and output, separately
for primary and secondary sources, which will improve the
measurement data for modelers. Status bits were added to identify
transactions which were active when either short on storage
(SHRTSTOR) or maximum task (MAXTASK) conditions occurred so that these
potentially delayed transactions can be excluded from response time
analysis. Program storage used was also added, and now the CICS
Monitor Facility (CMF) data completely maps all data which was formerly
available in the Performance Analyzer II product for transaction
analysis. In the global system data (CICSYSTM) the amount of storage
deleted by dynamic compression each interval (PROGCOMP) is provided.
The shock of 1.6.1 was massive, as the CMF version number was not
changed between 1.6.0 and 1.6.1. The MXG Code determines the
release by the order and number of variables in the record. MXG
Version ONE supported CICS 1.6.1.
Copyright (c) 1984,1985,1986,1987 Merrill Consultants, Dallas Texas
MXG is a registered trademark of Merrill Consultants
c. MVS/XA Version 1.2: The region requested was increased from two
to four bytes, and is now in variable REGREQST. Its predecessor,
PVTAREA, will still exist but will have zero value. MXG Version ONE
supports MVS/XA Version 1.2.
3. IBM APARs/PTFs Affecting Performance or Data
APAR 0Z75374 AND PTF UZ70647 - under MVS/XA at Release 2.1.2,
artificially high EXCP counts will be stored in Type 14 and 15 SMF
records. The PTF is on PUT tape 8404.
APAR OZ77211 and PTF UZ72177 - Causes invalid data in type 30. When
more than about 1407 DD segments occur, a second Type 30 record is
written with the additional DD segments. This second record seems
to have the device class, device type, and sometimes device number
invalid, frequently containing hex zeros. (An MXG message on the
log occurs if a DD segment contains nonzero EXCP count.)
APAR AZ74030 and AP97757 - problems with CICS CMF Type 110 SMF records
at two sites were reported to have gone away when these APARS were
installed. The first APAR deals with CMF problems when MRO is active,
and apparently required the second APAR as a prereq.
PTF PP11730 - CICS Monitor Facility (CMF) records have invalid second
header whenever the data in a series of Account Class segments exceeded
2048 bytes. This invalid segment is recognized and skipped in the MXG
APAR PP26559 and PTF UP43991 - CICS Monitor Facility under CICS 1.6.1
response times are all believed to be invalid without this PTF which
fixes incorrect values for terminal control wait time (WTTCIOTM). The
fix is on 8405.
PTF PP34408 - CICS Monitor Facility under CICS 1.6.1 with this PTF
reportedly causes bad data values.
4. MXG Software Information
a. Conversion Differences Between the Original Merrill's Guide and MXG
Considerable effort was taken to preserve the names of variables and
data sets in MXG so that users of the original Guide would not
have to alter their report programs. There are only a few areas in
which these changes might require recoding of your original report
i. TYPE71 RMF paging variables originally contained the total
page counts for the interval. As a result, report programs had to
divide the paging values by DURATM to calculate paging rates. MXG
corrected this original design error and calculates paging and
swapping variables as rates (pages per second).
ii. JOBS data set originally used variable JOB_TSO to
differentiate between jobs and TSO sessions. Now, with SMF data for
started tasks TYPETASK replaces JOB_TSO, and provides more
complete classification. TYPETASK exists in the STEPS, JOBS and
TYPE30_x data sets.
iii. For multi-processors, TYPE70 originally created
multiple observations for each Type 70 record, one observation
per "engine". MXG simplifies the data by creating a single
observation for each interval, which contains overall CPU busy
for all processors, as well as PCTCPBY0, 1, 2, and 3 for each
iv. Type 30 records originally created a single data set, TYPE30.
To reduce the disk space required by having all SAS variables for
every observation, MXG builds separate data sets (TYPE30_1,
TYPE30_4, TYPE30_5, etc.), keeping only the appropriate variables.
b. Specify SYNC in PARMLIB for RMF Synchronization
The RMFINTRV data set summarizes RMF data into a user-specified
interval. Records written at short intervals can be combined into
longer intervals for analysis. RMFINTRV requires that SYNC was
specified (the default is NOSYNC) in SYS1.PARMLIB. SYNC causes
RMF records to be synchronized with the wall clock, and must be
iv. Typographic corrections to the Book
p. 217. In code, replace RPTCICS (twice) with ANALCICS.
p. 309. In code, replace MGX (twice) with MXG.
p. 316. In 2nd paragraph of text, JCLTEXT should be JCLTEST.
p. 316. In last example, SOURCELIB should be SOURCLIB (twice).
p. 322. In 3rd line from bottom, TEXT DD should be TEST DD.
p. 371. Under SMF manual third line, the TNL is GN28 vice GN25
p. 373. 2nd entry, change Henning to Hemming.
p. 496. TIOESTTA bit 7 DASD, change dtaa to data.
p. 597. Penultimate line, change IMACINTY to IMACINTV.
p. 808. Text under CMD, change MG080CM to MG090CM.
p. 815. Remove line referring to JCLDBANL.
MXG NEWSLETTER VOLUME I NUMBER FOUR APRIL 1, 1985
Many users have directly contributed to these enhancements,
either by reporting problems, making suggestions, supplying raw SMF
records for code validation, or even by providing SAS code for
inclusion. These friends receive our sincere thanks, and their
contributions are identified in the member CHANGES. Additional
documentation often can be found in the comments at the beginning of the
MXG NEWSLETTER VOLUME I NUMBER FIVE JULY 5, 1985
2. IBM APARs/PTFs affecting performance or data.
a. High CPU utilization by CICS noted after installing CICS 1.6.1 can
result from two options. The VTAM High Performance Option (HPO)
defaults to off (NOHPO), which does cause excess CPU consumption.
Additionally, the FETRACE option always comes up active. FETRACE
can be temporarily disabled by the console command CSFE FETRACE,OFF
or permanently by APAR PP20143. This is probably common knowledge
among experienced CICS tuners.
b. SMF data destroyed at IPL. The SMF writer locates the last block
in the active SMF data set, but then overwrites it with the type 0
record. See APAR OZ85729.
c. SMF task ABENDs with 0C4 due to incorrect handling of a cross memor
post. PTF UZ79645.
d. DB2 data invalid. Three fields are overlaid. APAR is PP42547, the
fix is PTF UP59374.
3. Technical Information
a. The type 78 data on virtual storage usage in TYPE78VS and TYPE78PA
data sets is sampled at one tenth of the normal RMF sampling rate.
This is why variable NRSAMPLE in these data sets is one-tenth that
in other RMF data sets.
b. The Global Resource Serialization (GRS) address space requires a
significant amount of real memory unless GRS is placed in a control
performance group (by specifying it either in SYS1.PARMLIB or with
the PERFORM keyword).
MXG NEWSLETTER VOLUME I NUMBER SIX OCTOBER 25, 1985
I. MXG SOFTWARE INFORMATION
A. OVERVIEW OF ENHANCEMENTS IN MXG VERSION 3.
An enhancement to RMF required by MVS 2.1.3 or 3090's. Major items:
SWAP counts by type (SWAPUS, SWAPDW, etc) are now expanded with five
added variables (use of Extended Storage) for each swap type in TYPE71.
SU_SEC constant is now in the TYPE72 record, eliminating table lookup
IO measurement is different. The TYPE78CF data set, formerly only a
static tabulation of the configuration, now contains IO
utilization/queueing data for the CHPID. Two new data sets,
TYPE78CU and TYPE78IO now describe the utilization/queueing for the
CUHDR and IOP initiative queues. TYPE74 now contains three
components of DEVPNDTM. We plan a future enhancement to merge this IO
data for better analysis. The documentation for both type 71 and 78
records is seriously wrong in the SMF manual.
D. DOCUMENTATION OF CHANGES AND ENHANCEMENTS.
Changes and enhancements are documented in the Version 3 SOURCLIB
library. Member CHANGES of the library describes the changes and
enhancements in between the present and prior versions. Member
CHANGE02 describes the changes between Version 1 and Version 2.
Similarly, DOCVER02 and DOCVER03 members of the library contain the
documentation of the variables and data sets new with that
version. We plan to continue this philosophy and naming convention.
See section 4, below.
3. TECHNICAL NOTES
A. EXECUTION OF MVS UNDER VM CAUSES TYPE78 DATA TO BE MISSING.
To construct the configuration and recording of TYPE78 data, RMF must
read the IOCDS data set. Under VM, IOCDS is not available to MVS
and thus some type 78 records will not exist when MXG executes under
B. FIELD SIZES EXPANDED IN RMF.
MVS 2.1.3 shows some foresight for the future. Several formerly one
byte counters have been expanded to two bytes, suggesting that IBM
sees the need for more than 16 of them. Fields noted are number of
LCU's, LCUID, and several workload variables such as the
minimum and maximum MPL, Multi-Programming Level, and SRM values for
WEIGHT, AOBJ, DOBJ, and DFWK.
C. CONCATENATED %INCLUDED LIBRARIES.
If your source libraries have different blocksizes, you may create
trouble. SAS currently requires that concatenated source libraries
with dissimilar blocksizes which are read with %INCLUDE, must specify
DCB=BLKSIZE=largest on the first DD of the concatenation. If the
blocksize increases above the blocksize of the first DD, a loss of
source lines will occur, hopefully causing a syntax error.
E. TYPE74 VARIABLES CHANGED BETWEEN MVS/370 AND MVS/XA
Because device usage under MVS/XA is recorded differently, more
accurately, and by built-in hardware monitors, the MVS/370 variable
DEVBUSY in TYPE74 is missing under MVS/XA and the new variable PCTDVUSE
should be used. It provides a better measure of utilization - the
percentage of the interval when the volume was not available to another
A. ENHANCEMENTS TO EXISTING MXG DATA SETS
Support for the new restructured type 6 record from both JES2 and JES3.
See Washington Systems Center Flash bulletin WSCF-8518 for the PTFs
which create these new formats. New variables BINUSED, DUPLXUSE, and
SHEETPRN added for 3820 printer were added.
RMF 3.3 expanded storage adds several new variables:
PGMVTOEX PGMIEXAU dealing with page movement
AVLEXTMN,MX,AV for available extended storage
HIUICMN,MX,AV for HIUIC value
MIGAGEMN,MX,AV for Migration Age
EXTFRMIN, EXTFRMON for installed and online frames
There are now five new variables for each of the swap reason
PHYAUXxx Physical to Auxiliary
LOGEXTxx Logical to Extended
LOGAUXxx Logical to Auxiliary
PHYEXTxx Physical to Extended
EXTAUXxx Extended to Auxiliary
where xx is one of the SWAP reason codes.
Variables RESPAVG and RESPSTD, average and standard deviation of
response time per ended transaction were added stored by IBM in the
type 72 record, eliminating the table lookup, and more important,
eliminating the need for table maintenance every time you install a
Changes due to RMF 3.3 and enhancements
Three components of pending time, PCTDVPND, are now measured in
three new variables, PCTPNCUB PCTPNDEV PCTPNOTH are the percent of
pending time which is attributable to the Control Unit, Device, or
MXG variable OVERFLOW was split into two variables, OVFLOCON and
OVFLODIS for 308x. (Hardware counters in 309x and 4381 are
4-bytes, essentially eliminating the problem of overflow).
Omissions corrected and new variables AVG and INTRVAVG are created.
Enhancements due to RMF 3.3: This data set formerly contained
only the static data describing the physical configuration. With RMF
3.3, the rate at which this channel path (CHPID) was taken and the
percent when the Control Unit was busy and CHPID path busy are
recorded. As a result, TYPE78CF is now created by BUILDPDB. Variables
CHPIDTKN, PCTCUBSY and PCTPTHBY were created.
DB2 type 100 and 101 records introduced in Version 2 had logic
errors in processing IFCID and ASID segments of the type 100 subtype
0 (MXG dataset DB2STAT0) fixed. New variable QBACRIO, read requests,
found in DB2ACCT. See DOCVER03 for details.
Support for Release 4.3 of the TSO/MON Product. TSOMCALL variable
PANELCNT, SPF panel count was added. Seven new TSOMSYST variables
added, included the new estimated Network Delay Time per transaction,
B. NEW MXG DATA SETS
1. TYPE22_9 Extended Storage Configuration.
Like all type 22's, it describes the configuration. The subtype 9
describes how much extended storage is installed.
2. TYPE39 Network Logical Data Manager Release 3
Support for the Network Logical Data Manager, NLDM Release 3, which
creates type 39 SMF records is provided in the several TYPE39 data
sets described below. More information is described in NLDM
Operation, SC30-3165-2. Of particular interest is the capture by
NLDM of the response times captured by the 3274 control units.
TYPE39_1 - RTM Collection Record
TYPE39_2 - Session End Record
TYPE39_3 - Session Start Record
TYPE39_4 - Accounting and Availability Data Record
TYPE39_5 - Combined Record
TYPE39_6 - Bind Failure Record
TYPE39EL - Route Element Table
3. TYPE78CU RMF 3.3 Control Unit - Header Queue, CUHDR
Describes the Control Unit - Header Queue, CUHDR, with queue length and
rate. When all paths to the subchannel are busy for an LCU, and at
least one path to the control unit is busy, IOP places an IO on the
4. TYPE78IO RMF 3.3 IOP Initiative Queue.
Describes the Input/Output Processor, IOP, initiative queue statistics.
5. TYPEDISO DISOSS Accounting Log Record.
Data set DISOACCT (name not TYPE.... because it is not created from
an SMF record) is created from the DISOSS log. See technical
note, above, or DOCVER03.
D. ADDITIONAL COMMENTS
1. BUILDPDB and BUILDPD3
Several enhancements made to BUILDPDB/BUILDPD3.
a. ABEND and CONDCODE in the JOBS data set is now the job completion
status for the last job termination (subtype 5) record. Formerly, it
was the last step termination and was misleading (i.e., ABEND would
be FLUSH in PDB.JOBS for a job whose last step was flushed due to the
ABEND in a prior step).
b. EXCP and IOTM for tapes 3420 and 3480 added.
c. Data sets TYPE78CF, TYPE78CU, and TYPE78IO are now created in the
PDB. All three are physically small.
New format, MG078CV. converts byte values to Kilo or Mega bytes, with
K or M suffix. Makes the data in TYPE78VS (virtual storage) easily
read. Try it with any variable in our data sets. We also use it in
TYPE30 data sets.
LABELS were enhanced to reflect that many variables are rates per
second. Labels for variables ending with SWAP, TRAN, EXCP, IOTM, ACTV,
RESD, and SERV now have "per second" in their label.
The calculation of ALOCTIME and LOADTIME in TYPE30_4 were corrected
(as had been done in TYPE434) to add a day if the job allocation/load
time of day was less than the initiate time of day, i.e., for jobs
which initiated today and allocated/loaded after midnight.
7. TYPE30_4, TYPE30_5, and TYPE434
Return code values are limited by IBM to 4096, but with a two byte
field, some accounting packages use the upper bits for their own
purposes. This caused return codes greater than 4096. The MOD
function now returns only the real return code set by the programmer.
The _VAR70 ARRAY in processing type 70 records was replaced with
direct assignment after a European customer first noted a significant
increase in CPU time of MXG (140 seconds) when compared to
the original code (84 seconds). The ARRAY was more CPU costly than
direct assignments, and had been used only as shorthand.
9. X-named members
All MXG modules beginning with the letter X are "extra's". Their
only documentation is the member itself, and these members are not
supported at the present time. They may not necessarily work. Some
might eventually be upgraded to supported.
This module partially decodes SMF record types 0, 1, 4, 5, 7, 12, 20,
34, and 35 from the OS/VS1 operating system. It now executes, after
removal of an unnecessary %INCLUDE and proper calculation of NUMDD
to make it usable for VS1 data records. If there is enough demand, we
will upgrade it with labels, re-structure it and enhance it into
a supported member. The variables created are already described in
the corresponding sections of Chapter 40.
* This product could not exist as it is without the wonderful *
* user community with whom we work. Your calls and letters, *
* sometimes with dumps and sometimes with data tapes, provide *
* us with opportunities to detect our errors and to protect our *
* code (and your sleep!) from interruptions. *
MXG NEWSLETTER VOLUME I NUMBER EIGHT MAY 23, 1986
Major items in Version 4.1, available upon request.
1.New module IMACPTF. Presently used only for CICS SMF records,
this new maintenance feature makes it easier when you install
IBM PTFs. If we knew about the PTF at shipment time, it will be
in this module, and by enabling the PTF MACRO in this module, you
enable MXG code to process the changes.
ALWAYS BROWSE IMACPTF with each new tape.
2.Support for CICS 1.7 has been added to VMAC110 code. Note that
new program UTILCICS is usable for either 1.6 and 1.7 to help
decifer what PTFs are installed, so you can update IMACPTF.
3.MVS 2.1.5/RMF 3.4 support. This includes all changes in the MXG
Newsletter number 7 and later. This also corrects several minor
errors in data sets new with RMF 3.3.
4.PRE-RELEASE MXG IMSLOG CODE enhancement (TYPEIMS). See member.
5.Revised ANALTAPE routine.
6.MODEL204 accounting information defaults on (EXM24ACT).
7.New member INSTALL with instructions for installation of MXG,
both new and replacement, as well as under MVS or VM. Please
use this member to install this version, and advise if the
instructions are not clear.
8.PRE-RELEASE processing of VM Monitor records directly with SAS
CMS Maintenance Release - not completely validated (TYPEVMON).
9.NODUP option implemented in BUILDPDB/3 to remove duplicate SMF
input data from PDB data sets. (Requires enabling after you have
installed the pre-requisite SAS maintenance release).
10.Too much more to list here. Read member CHANGES.
The MXG SUPPLEMENT will be automatically sent to supported sites,
as a special edition of the MXG Newsletter. Future Newsletters
will be in the same font and size to be inserted in binder for
the MXG SUPPLEMENT.
====THIS IS THE LAST LINE OF MEMBER NEWSLETTERS=================