* copyright (C) 1984-2019 merrill consultants dallas texas usa

Change 05.154 IMS Log variables DESTLINE and DESTTERM should have

Nov 18, 1987 been read as PIB4. vice PIB2., with HEX8. format.

VMACIMS

Thanks to Freddy Simberg, PK Banken Stockholm, SWEDEN.

Change 05.153 Several changes to Landmark's MONITOR for CICS:

Nov 18, 1987 1.Landmark's MONITOR summary history file records caused

TYPEMONI "Invalid data for TASKNR" because summary record has

nulls (hex 0) value. Error message eliminated and the

variable MONIRECD now correctly contains 'H' for the

History (Summary) observations and 'D' for Detail.

2.Observations in MONIDBDS and MONIUSER will now occur

only if the segment actually contains data. FATNUM

counts DBD segments in record, not the number which

contain data, and UTNUM counts the user segments, but

not the number which contain data.

3.Support for Landmark's SPECIAL 78 optional zap to

TMON Version 7.0 creates CICS 1.7 variables UOWID,

UOWTIME, USER, and NETSNAME in MONITASK data set.

Supplement pages 170-171 describe these variables.

UOWTIME is the decoded time stamp from first 6 bytes

of UOWID). Variable SYSTEM was also added to both

MONITASK and MONISYST data sets with this zap.

4.Initialization of all character values set by bit

testing and SKIP logic for any new data was added

5.Variable DURATM in MONITASK was not calculated for

the first interval during re-processing with DIF().

Two elegant algorithms were provided by two users,

and the shorter one (using POINT= to read ahead to

the next record) was chosen.

Thanks to Kathy Manos, The Stroh Brewery Co, USA.

Thanks to Robert Lewis, Landmark Systems Corporation, USA.

Thanks to Chuck Rexroad, Perpetual Savings Bank, USA.

Thanks to Simon Hendy, British Telecom, ENGLAND.

Thanks to Steve Morton, SAS Institute, ENGLAND.

Change 05.152 The CPU-specific variables PCTCPBYx "defaulted" to

Nov 16, 1987 100% busy for offline CPUs. This had minor impact,

VMAC7072 since variable NRCPUS showed how many were online,

amd also since CAIx flags each CPU as to on/offline

status. Furthermore, PCTCPUBY always accurately had

the true percent of online cpus time which was busy.

Nevertheless, the misleading 100% for PCTCPBYx has

been replaced with a missing value for CPUs which

were not online for the entire interval. (With this

change, the MVS/370 record logic was changed to use

the more robust MVS/XA technique.)

Thanks to Stephen Geard, Commonwealth Bank of Australia, AUSTRALIA.

Change 05.151 I finally accept the need to use the new SAS Macro

Nov 16, 1987 Facility. This change establishes naming conventions

for MXG members. The %MACRO facility works best if

the macro name is the same as the member name (then

AUTOCALL can be used). To facilitate this feature,

MXG will hereafter use names starting with VMXG....

TYPEVTOC for both the member and macro name of MXG-supplied

VMXGVTOC new style macros. To balance this "theft", MXG will

hereafter not use member names starting with USR....

or USER.... - these will always be available to you.

Thanks to Bill Gibson, SAS Institute, AUSTRALIA.

who's comment on "using VMACVTOC (now VMXGVTOC) to graze the disk

farm" was under study when this choice was made.
Change 05.150 IMACFILE, taken after an SMF record has been read by

Nov 16, 1987 any MXG code, is self-documented, and is used for many

IMACFILE purposes, including deleting bad SMF records, printing

VMACSMF a "defective" SMF record, etc. IMACFILE is invoked by

VMACSMF. This change "cleans up" comments in VMACSMF,

and now new comments in IMACFILE identify all of the

variables which exist when IMACFILE is taken.

Thanks to Jim Gilbert, Texas Utilities, USA.

Change 05.149 Variable ZDATE was added to VMONUSRS, the one-per-user

Nov 16, 1987 summary data set. This provides the date of MXG run,

VMACVMON not the processing date of the data. VMONUSRS is the

summary of all observations in VMONUSRD (which could

be daily, weekly, etc.). Tailored summary (by hour, by

day, etc.) can be done in your code from VMONUSRD.

Thanks to Jim Gilbert, Texas Utilities, USA.
Change 05.148 CICS 1.7 UOWTIME was not properly decoded in the pre

Nov 16, 1987 releases of MXG for terminal tasks. Now it is. (Note

VMAC110 that UOWID, the 8-byte character string by which MRO

transactions can be collected was always correct; only

the extraction of its timestamp was in error).

Thanks to Bernadette Young, First Virginia, USA.

Change 05.147 MVS 2.2's catalog address space creates a type 30

Nov 16, 1987 interval record with JOB=IEESYSAS & TYPETASK=CATALOG

VMAC30 which caused "Invalid JESNR" message (no error set).

This change tests for IEESYSAS (system address space)

job name to avoid missing JES number for address

spaces which start before SMF is started.

Thanks to Jim Gilbert, Texas Utilities, USA.
Change 05.146 Variable DEN (tape density) in TYPE1415 data was not

Nov 13, 1987 decoded for 3480 (density=38000) devices. Now it is.

VMAC1415

Thanks to Chuck Hopf, Dean Witter Reynolds, USA.

Change 05.145 Variables ROSJOB, ROSTIME, SMFTIME and SYSTEM only are

Nov 13, 1987 in ROSCOE 5.5 and later. The ANALRRTM reports, however

ANALRRTM used these variables, which were not defined in this

VMACROSC member (used only for ROSCOE 5.4A and earlier). This

VMACRRTM change initializes these four variables to blanks or

missing values so that ANALRRTM will work with either

ROSCOE 5.4, 5.4A, or 5.5 RRTM data.

Variable ZDATE was added to all ROSCOE data sets too,

and the code was prettied up.

Thanks to Rich Surgener, Security National Bank, USA.

Thanks to Jim Gilbert, Texas Utilities, USA.
Change 05.144 Change 5.44 should have also applied to ANALDB2. The

Nov 13, 1987 variable QXQUERY is now QXSELECT and QXALTBF is now

ANALDB2 QXFETCH in lines 331 and 354 respectively.

Thanks to David Daner, Sun Company, USA.

----Changes thru 5.143 as of Oct 22, 1987 were included in MXG 5.3a--
Change 05.143 Two syntax errors slipped into 5.3. prerelease tape.

Oct 22, 1987 1. Added semicolon to the end of line 2182.1 in member

VMACVMON VMACVMON.

VMACCIMS 2. Added DBDNAME $8. to line 180 (before semicolon)

in member VMACCIMS.
----Changes thru 5.142 as of Oct 15, 1987 were included in MXG 5.3---
Change 05.142 ROSCOE 5.5 support, including the AUDIT record and the

Oct 14, 1987 RRTM (response time) data, has now been added. This

TYPEROSC changes most of the ROSCOE members, and adds several

et. all. new data sets from accounting data. The RRTM records

are now a subtype of the SMF record, and the RRTM data

sets are now built by TYPEROSC. (Member TYPERRTM is

still in MXG if you are not yet on ROSCOE 5.5). Data

for all 5.5 records has been tested, except for the

5.5 Audit record (but that's not likely to cause you

any problem - ADR doesn't document how to enable that

new record, though they will tell you how by phone!)
Change 05.141 Boole & Babbage's IMF (formerly Control/IMS)

Oct 14, 1987 optionally will capture I/O at the DDNAME within

VMACCIMS DBDNAME for those DBDNAMEs with multiple DDNAMES. This

change adds the new variable DDNAME to the CIMSDBDS

data set. If the DDNAME is not blank, this is a DD

entry, otherwise it's a DBD.

Thanks to Ron Root, Sun Company, USA.
Change 05.140 The subtype 1 IDMS Performance Monitor record from

Oct 14, 1987 the new PMAM available in IDMS 10.1 is being created

VMACRTE in error. This change skips over the subtype 1 PMAM

data until Cullinet provides a modification (accepted

but not yet available) to correct their error.

Thanks to Rodney L. Reisch, General Electric Silicon Division, USA.

Change 05.139 The BDT (Bulk Data Transfer) record has now been

Oct 14, 1987 tested with real type 59 SMF records. Two changes were

VMAC59 required to VMAC59: line 1620, TRANTYID replaced

SMF59TID in the test, and line 1670's SMF59OFG field

is now input with a $CHAR1 instead of the incorrect

$CHAR4.

Thanks to Doris Burdick, U.S. Department of Defence, USA.
Change 05.138 The IMS log processing with TYPEIMS from MXG 4.4

Oct 11, 1987 fails. Most transactions were being thrown away

IMACIMS because of the changes in the IMS records introduced

TYPEIMS in 1.3 and 2.1. The pre-releases 5.1 and 5.2 modified

VMACIMS the MXG 4.4 code to stop deleting records, and did

correctly capture all resource data, but introduced

occasional transposition of the LTERM and TRANSACT

values. Most of all, I just did not understand the

logic of the donated code. This change marks the first

IMS log processing code which I actually wrote and

understand. The only original code in TYPEIMS/VMACIMS

now is the merge of the 07-08 log records. Larry

Alexander's fine IMS paper and 2.1 data were used to

now properly assemble all simple IMS transactions

perfectly. The existence of an 07/08 causes an

observation in IMSTRAN to be created, and if an 01

record was found, the message queue records (01,03,31,

35, and 36 are then collected under the transaction's

LTS (line, terminal, sequence) by their message queue

DRRN (dataset and record number) and the time stamps

(ARRVTIME,ENQTIME,GUTIME,STRTTIME,TMCNTIME, GUOUTIME,

ENDTIME and DEQTIME) are captured. Variables INBITS

and INCNTS identify which and how many IMS log records

were collected for this observation in IMS tran. This

code has processed many IMS log records, and seems to

be correct in assembling matching sequences. For some

very complicated sequences (those which do create an

01 for each 07, for example) will be fully assembled

into IMSTRAN for resource measurement and transaction
counting. The worst that can happen is that some of

the time stamps may be missing if MXG can't find the

message records corresponding to the 07 and 08 log

records. I am still investigating WFI (wait for

input) transactions, and may provide additional code

in the future to re-examine IMSTRAN see if the IMS

response for these unique transactions can also be

measured. In summary, the 5.3 IMS code is thought to

be completely accurate for resources and counting, and

either captures the correct response time, or sets the

response time missing. It's not done, but it's so

much better than anything before that it should be

used now for any IMS log processing.

Thanks to John D. Pike, American Airlines, USA.

Thanks to John Lemkelde, Pennsylvania Blue Shield, USA.

Thanks to Finn Poulsen, DANFOSS A/S, DENMARK.

Change 05.137 A syntax error resulted from a missing AXIS2 statement

Oct 11, 1987 in a PROC GPLOT. New line 370.1 should be:

GRAFRMFI AXIS2 COLOR=WHITE LABEL=('LOGICAL SWAP')

Thanks to John Potter, Mitsubishi Electronics, USA.

Change 05.136 ROSCOE 5.5 TERMNAME variable is now only in the signon

Oct 4, 1987 record, but 5.4 had it also in the logoff record. The

ANALROSC result was a blank value for TERMNAME. To correct,

change line 8 to now read:

PROC SORT DATA=ROSCOFF (DROP=TERMNAME);

Thanks to Richard Olimpio, O. M. Scott, USA.

Change 05.135 RMF Cache Reporter records which have only one storage

Oct 4, 1987 director segment caused INPUT STATEMENT EXCEEDED

VMACACHE RECORD LENGTH, "STOPOVER" condition. (All data tested

had two!) These changes were made:

Insert new lines:

@15+OFFSTAT GSLEN PIB2. 237.1

@; IF GSLEN=80 THEN INPUT 245.1

(SD2ID=SD2IDCU OR GSLEN=40) ); 258.1

Remove the "SD2ID=SD2IDCU);" at end of line 258.

Thanks to Benny J. Arnold, Kansas City Computer Center, USA.

Thanks to Don Tuttle, Kansas City Computer Center, USA.
Change 05.134 If you wanted to build the PDB data sets and only want

Oct 4, 1987 to spin once (for example, building from a monthly SMF

BUILDPDB file), you could not. Setting IMACSPIN to 1 caused the

BUILDPD3 un-matched records to be spun twice, since _SPINCNT

was compared with "1" in line 375/385 (BUILDPDB/3).

Changing the "1" to "0.5" solves this inconsistency.

Thanks to Bill Gibson, SAS Institute, AUSTRALIA.
Change 05.133 1.VM/Monitor USER records contain no logon flag. MXG

Sep 23, 1987 can recognize a logon if any of the DIF()'d counters

Nov 29, 1987 went negative, but that is not always the case. Now,

ANALVMON MXG compares the delta minutes since the last record

IMACVMON for this user with a new macro (_INTRVL, in IMACVMON)

VMACVMON in which you specify your expected interval duration.

If more than two expected intervals have been missed,

MXG assumes a LOGON and reinitializes. (Note that if

suspension occurs, records can be skipped.) The actual

comparison is IF DIF(STARTIME) GT 2 * _INTRVL THEN ...

The default value for _INTRVL is 1 minute. If you are

writing 10 minute intervals and you do not change the

value of _INTRVL in IMACVMON, all data will be zeros.

2.Twenty-one VM/Monitor durations (CPU times, Page and

I/O waits, etc.) are in counters which decrement from

a large positive number down to zero. MXG did not get

the correct value for the zero-crossing interval. The

problem only occurs when the system stays up for more

than a day (after 38 hours at 50% IDLETM, it resets).

The error caused the zero-crossing variable to contain

a very large negative value for that one interval. The

decrementing counters are identified by the syntax of

y=-DIF(x)/4096, which calculates the negative delta of

counter x into variable y. This change inserts a new

test for negative y following each -DIF() to correct:
y=-DIF(x)/4096;

IF . LT y LT 0 THEN y=y + 1E-6*16**9;

This fix was revised Nov 29, when IBM Level 2 made

it clear: six bytes of FFFFFFFFFFFF is stored. MXG

5.3 incorrectly used just 16**9, the 5.4 temporary

circumvention used x/4096 pending this truth. Note

that computerese 1E-6*16**9 is simply 16 to the 9th

divided by a million, but the 68719.476736 decimal

equivalent leaves me cold. (Six bytes of FF, INPUT

as PIB6.6 is decimal 281474976; divided by 4096 to

get seconds, becomes 16 to the 9th over a million!)
3.I/O counts in VMONDEV are accumulated by DEVADDR since

last IPL, and MXG failed to de-accumulate the variable

CNTVIOCT. Macro _ANALDEV is now defined in VMACVMON

to sort and correctly calculate CNTVIOCT into VMONDEV.

(Note for recipients of pre-releases 5.1 or 5.2: This

de-accumulation WAS correct in the VMPDB.DEV data set

built by ANALVMOS. Now, the de-accumulation is done

in building VMONDEV). This change could impact the

compatibility of MXG. If you used VMONDEV from MXG 4.4

you probably did your own de-accumulation, and now you

will need to remove that de-accumulation.

Thanks to George Moskwa, Cole National, USA.

Change 05.132 If SAS OPTION NOTEXT82 is in effect, a syntax error in

Sep 23, 1987 creating the temporary $TCLASS reporting format in one

ANALCICS of these sample CICS reports. The right hand side of

the format values must be enclosed in single quotes,

or OPTIONS TEXT82 can be specified.(Our testing now

specifies NOTEXT82!)

Change 05.131 Variables INCONT, LOGONID, MSGLEN and ORGENT in the

Sep 23, 1987 TYPEIMS data set (from IMS Log records), which exist

VMACIMS in the RACF segment if you have RACF or ACF2 with IMS,

but were not in the KEEP list for IMS0103 data set.

As a result, they were blank. (FYI, they are defined

in member EXITRACF.)

Thanks to John Lemkelde, Pennsylvania Blue Cross, USA.
Change 05.130 TSO/MON 5.1 support has now been validated with data

Sep 21, 1987 containing the optional ACCOUNTing fields. Without

VMACTSOM this change, STOPOVER abend occurred if the additional

optional data was present in the record.

Line 2417 changed to INPUT NRACCTFL PIB1. @;

Lines 24192 and 24193 were deleted. (MINUS8= and INPUT)

Thanks to John Leath, Fleet Information, USA.
Change 05.129 The MONIUSER data set from Landmark's MONITOR has now

Sep 21, 1987 been validated. CLOCKTM is now calculated by 1E4, and

TYPEMONI the code was restructured to properly initialize.

Thanks to Chuck Comstock, Hallmark Cards, USA.

Change 05.128 Jobs with JCL errors in PDB.JOBS had blank value for

Sep 18, 1987 TYPETASK because it was not in KEEP list for the type

BUILDPDB 26 data. TYPETASK was added to _PDB_26 definition.

BUILDPD3

Thanks to Lee Salley, Westinghouse, USA.
Change 05.127 Several minor changes.

Sep 16, 1987 1.ANALAUDT and ANALDB2R were restructured to contain

ANALAUDT both JCL and SAS code. With the SAS JCLEXCL option:

ANALCICS %INCLUDE SOURCLIB(ANALDB2R)/JCLEXCL;

ANALDB2R SAS will skip over the JCL and execute the SAS code.

EXITMONI Alternatively, the JCL can be edited and submitted to

produce the report from either an MXG PDB library or

directly from SMF data. Both ANALAUDT and ANALDB2R now

permit date selection via the SYSPARM parameter on the

EXEC statement, as described in the comments. MXG will

use this pattern in future ANAL.... members.

2.ANALCICS PROC FORMAT at line 465 needed quotes on the

right hand side if SAS OPTION NOTEXT82 is specified.

3.EXITMONI installation instructions were improved.

Change 05.126 MXG Format $MGIMFFP was defined in lines 411-414 of

Sep 14, 1987 this member in MXG 4.4, but is not in MXG 5.2 member.

FORMATS If you use the first step of JCLTEST to add formats to

your existing SASLIB data set, you will not see this

error. However, if you create a new SASLIB using the

5.2 FORMATS member, the format will not exist. (Since

I tested by adding the new formats, my testing failed

to catch the error - guess who's testing philosophy

has been changed!). To correct the 5.2 FORMATS member,

copy lines 411-414 of 4.4 FORMATS member after line

615 of 5.2 FORMATS member.
----Changes thru 5.125 as of Sep 10, 1987 were included in MXG 5.2---
Change 05.125 JES2 sites using NJE will now find the PDB.NJEPURGE

Sep 10, 1987 data set automatically built by BUILDPDB will contain

BUILDPDB the non-execution purge records for NJE jobs. All SR

(SYSOUT Receiver) purge records and JR (job transmit)

purge records which do not represent actual execution

will be in this PDB data set. (Formerly, these purge

records were deleted by BUILDPDB). This data set may

be useful in tracking NJE job transmission input

delays as well as NJE SYSOUT delays. Additional work

in this NJE measurement area can be expected.

Thanks to Gary Zolweg, National Semiconductor, USA.
Change 05.124 Support for ROSCOE 5.5 Accounting records was added.

Sep 10, 1987 The ROSCOE Response Time Monitor (RRTM) data is now a

EXROSALL subtype of the 5.5 account records written to SMF.

EXROSATT Several new ROS..... data sets are created from the

EXROSCON new ROSCOE 5.5 account records. On October 11, 1987

EXROSDSF (in MXG 5.3), RRTM record support for ROSCOE 5.5 was

EXROSHEX added to these members, and the RRTM data sets are now

EXROSLIB created with TYPE/VMACROSC. The AUDIT record (new in

EXROSPUR ROSCOE 5.5) support was also added, but no audit data

EXROSSHU records have been available to test yet.

EXROSSOR

EXROSVPE

IMACROSC

VMACROSC
Change 05.123 VM Monitor summarization & documentation were cleaned

Sep 10, 1987 up in testing 5.1. The SEEKRPRT has been added and the

ANALVMOS documentation improved. VM Execs referenced in the MXG

DOCVMOS supplement now exist. EXECTEST is the VM equivalent of

EXECPDB JCLTEST, EXECPDB tests the MVS PDB building under VM,

EXECTEST and EXECEMAC provides an easy way to edit MACLIBs.

EXECEMAC
Change 05.122 Installation modifications to type 26 records caused

Sep 9, 1987 MXG problems because section lengths were not used to

VMAC26J2 +SKIP over new data. This change redesigns TYPE26J2 to

fully use the section lengths for immunity to change.

Thanks to Rob Clairmont, The Laurier Group Ottawa, CANADA.

Change 05.121 Several ANAL.... report members created unnecessary

Sep 9, 1987 messages (not referenced, uninitialized, etc.) which

ANALPROG were irritating but not critical errors to new users.

ANALESV

ANALTAPE

Thanks to Kathleen A. Morrison, Hit or Miss Inc, USA.

Change 05.120 The IDMS Performance Monitor (Change 5.95) did not

Sep 9, 1987 process the second subtype 1 record, which is now

VMACRTE recognized and creates the IDMSTASK data set.

Thanks to Rodney L. Reisch, General Electric Silicon Division, USA.

Change 05.119 Variables EXCPTODD/EXCPNODD and IOTMTODD/IOTMNODD in

Sep 9, 1987 TYPE30 data sets are missing if the program had no DD

VMAC30 cards. Uncommon, it does happen. MXG moved several

lines outside the NUMDD=0 do group to set these four

variables zero rather than missing in this instance.

Thanks to Kenneth D. Jones, Maritime Telephone and Telegraph, CANADA.

Change 05.118 Only for the TYPE74 data set and only under MVS/370,

Sep 9, 1987 the variable DEVNR (the MVS/XA and BUILDPDB/3 expected

VMAC74 variable name for the old UNITADR MVS/370 variable) is

missing. Correct for MVS/370 with the new line:

DEVNR=UNITADR; 172.1

to store unit address in DEVNR. MVS/370 sites would be

wise to position for MVS/XA by changing UNITADR to

DEVNR now in all programs using MXG data sets.UNITADR

does not exist under MVS/XA, DEVNR replaced it.

Thanks to Bob Rutledge, Sherwin Williams Paint, USA.

Change 05.117 VM Account support for the type 08 (user disconnect

Sep 9, 1987 or log off) account card record was added, creating

TYPEVM new data set VMLOGOFF. Variable LUNAME was also added

EXVMLOFF to the VMDISK, VMLINK, VMLOGON and VMLOGOFF data sets.

FORMATS LUNAME will contain the SNA (VTAM) terminal name for

SNA terminals, useful for identification of terminals

being used by VM users. In VMDEVICE data set, variable

DEVICECL is now decoded by new format $MGVMADV, which

permitted reducing DEVICECL from 8 to 1 byte.
Change 05.116 The format used in the PUT function at line 46 was

Sep 9, 1987 corrected from WEEKDAY3. to WEEKDATE3. to create the

MONTHBLD name of the day of the week as a character variable.

Thanks to Barry Lampkin, Blue Cross of Mass, USA.

Change 05.115 MXG error messages VMAC110.2 and .3 were revised to be

Sep 3, 1987 more descriptive when MXG detects a change in record

VMAC110 format due to CICS maintenance.