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

single hex digit of the actual drive within that CU on

which the tape volume being dismounted was created.

For 3420s it contains the creation tape drive serial.

UCBTYPEs length is increased to 5 to avoid truncation.

Member ANALESV was also modified to print 3480 values;

those report changes are not given here.

Thanks to Norbert Korsche, OMV-AG, AUSTRIA.

The following change to VMAC21 is only needed if you are interested

in analyzing 3480 tape errors. It will be automatically installed in

MXG Version 5, as are the changes to ANALESV.
The change alters four existing lines and inserts twelve new lines.

The line numbers of the changes/inserts are shown below. Inserts have

a period after the line number they are to be inserted after. Changed

lines have no period.

Source statement Line number
DCBOFLG DENSITY DEVICE DEVNR 0007

OPEN 0010.1

TAPCUSER TEMP.... (rest unchanged) 0013

DEVICE ='DEVICE*TYPE ' 0027.1

OPEN ='TYPE*OF*OPEN ' 0032.1

LENGTH DEFAULT=4 UCBTYPE 5 SMFTIME 8; 0048

TAPCUSER $HEX6. 0051.1

@27+OFFSMF DEVCLASS PIB1. 0055.1

@28+OFFSMF DEVTYPE PIB1. 0055.2

%%INCLUDE SOURCLIB(VMACUCB); 0067.1

IF LENGTH GE 58 THEN DO; 0068

@44+OFFSMF TAPCUSER $CHAR3. 0070

IF DCBOFLG='0.......'B THEN OPEN='INPUT '; 0076.1

ELSE IF DCBOFLG='1.......'B THEN OPEN='OUTPUT '; 0076.2

BYTEREAD=4096*BYTEREAD; 0076.3

BYTEWRIT=4096*BYTEWRIT; 0076.4

END; 0076.5

Original lines 86-87 MOVEed to new lines 76.3 and 76.4

Change 05.20 The "Total Recorded CICS CPU Time", created only in

Dec 23, 1986 MXG reports in ANALCICS, is incorrect. As described in

Jan 26, 1987 the MXG supplement, the total CPU time recorded in the

ANALCICS CICSYSTM data set should have been(line 88):

CPUTOTTM = SUM(CPUSRBTM,CPUTCBTM);

NOTE: This only affects ANALCICS reports, not the MXG

CICSYSTM data set. See Chapter 21 in MXG Supplement.

Thanks to Beth Asbury, Prudential Home Mortgage, USA.

Change 05.19 PVTBOT, PVTTOP, and REGREQST were converted to bytes

Dec 23, 1986 (their value was multiplied by 1024), and they are now

VMAC30 converted by MXG format MG078CV., which prints bytes

as K or M, depending on magnitude. This was done to

See also make PVTBOT, PVTTOP and REQREQST consistent with the

Change 6.13 other virtual storage measures in TYPE30 data sets,

LSQSZHI/LOW, PVTSZHI/LOW and USRSZHI/LOW.

Thanks to Norbert Riehout, Great Western, USA.

Change 05.18 DL/I Counter decoding was added for CICS 1.7. The same

Dec 23, 1986 four clocks and nine counters are decoded, but their

IMACICDL order in 1.7 is reversed from 1.6: the four clocks are

first, followed by the nine DLI counters, Also, the

four PIB2. count variables for the four clocks must be

changed to PIB4. Finally, the test for length of data

must be changed from 60 to 68. Copy the 1.6 decode to

the 1.7 logic area in IMACICDL and make these changes.

Thanks to Thomas Victor, Kaiser Data Center, USA.
Change 05.17 These new members provided by Gothaer Vershicherung in

Dec 15, 1986 1.TYPECACH member supports the 3880 Cache DASD RMF

TYPECACH Reporter FDP SMF records in two datasets for the model

VMACACHE 11/21 and 13/23 cache.

VMAC60 2.Member VMAC60 has been updated to decode the VSAM VVR

VMAC60 fields, although none of these variables are labeled

SYSLOGJ3 and some will need MXG formats.

UTILSPAC 3.SYSLOGJ3 processes the JES3 SYSLOG. Worth reading.

4.UTILSPAC collects DASD size (space) info from CVAF

VTOC's.

Change 05.16 Decode of MONIFLG2 is incorrect. Replace lines 428-430

Nov 24, 1986 with:

TYPEMONI '...1....' THEN MONIATI='Y';/*ATI*TASK*/

'....1...' THEN MONSWAIT='Y';/*VSAM*STRING*WAIT*/

'.....1..' THEN MONBWAIT='Y';/*VSAM SHARED*BUFFER*WAIT*/

Change MONDMONI to MONIATI in line 45, and replace 121

(MONDMONI label) with MONIATI='ATI*TASK'
Change 05.15 Replace lines 101-105 (HS,MS,SS,TS input) with:

Nov 24, 1986 STARTIME PIB4.2

TYPEDISO and remove line 126 (STARTIME=HMS...)
Change 05.14 Additional test to prevent STOPOVER abend with 1.6.1.

Nov 24, 1986 Add to present criteria for VMAC110.1 error message:

VMAC110 OR (CONNBYTE+TEMPBYTE GT TEMPLEN AND MCTSSDID NE 0)

Thanks to Chuck Hopf, Dean Witter Reynolds, USA.

Change 05.13 Concatenated multiple VM Monitor files caused negative

Nov 24, 1986 CPU values at beginning of each group of monitor data.

VMACVMON To reset INTRVCNT when new BEGIN monitor (CL 0 CD 97)

record is encountered, insert new line:

ELSE INTRVCNT=0; 281.1

after line 281.(Change was revised Feb 16, 1987.)

Thanks to Tim O'Rourke, Bankers Trust NYC, USA.
Change 05.12 IRC Wait time in CICS 1.7 is located after SUSPNDCN.

VMAC110 a. Change line 678: WTIRIOCN PIB2. to WTIRIOCN PIB4.

Dec 15, 1986 b. Move lines 676 thru 679 (IF _PP43887 THEN thru @;)

immediately after line 688.

c. The actual CICS 1.7 APARS which add WTIRIOTM/CN are

APARs PP51612 and PP56560. To enable MXG for these

PTFs, however, you must enable the macro named

_PP43887 (the CICS 1.6 number) in IMACPTF.

d. UTILCICS will not detect the existence of the IRC

data fields. However, if IRIOWTT appears under the

column heading of CMODNAME on the dictionary print

from UTILCICS, then that CICS 1.7 APPLID must have

_PP43887 enabled in IMACPTF. UTILCICS will be

modified to check CICSDI17 to report this PTF.

Thanks to John Loo, Fluor, USA.

Thanks to Joseph J. Faska, Chemical Bank, USA.

Change 05.11 MXG did not correctly initialize all variables. Under

Oct 31, 1986 some circumstances, character flags (eg, VARY, ONLINE)

VMAC73 could be erroneously carried to subsequent devices.

VMAC74

VMAC78

Thanks to Paul Larson, Marine Bank Services, USA.

Change 05.10 A bad CICS 1.6 data block could cause strange errors.

Oct 31, 1986 Further testing added to delete bad CICS blocks.

UTILCICS

Thanks to Kirby Hindin, Allnet Communication Services, USA.

Change 05.9 Would not read VSAM SMF records correct. +OFFSET was

Oct 31, 1986 missing in lines 303,314, and 334.

VMAC1415

Thanks to Chuck Hopf, Dean Witter Reynolds, USA.

Change 05.8 Full duplex line utilization could exceed 100%, due to

Oct 22, 1986 CHRSPEED being the MAX of SEND or RECEIVE line speed.

VMAC38 HDX line has only SEND or RECEIVE non-zero.

To fix, change the CHRSPEED=MAX() to CHRSPEED=SUM().

Change LSPEED=MAX() to LSPEED=SUM().

Thanks to Marcia Geyer, City of New York FISA, USA.

Change 05.7 Syntax error due to line dropped after line 3200.

Oct 22, 1986 SET TYPE0; added following DATA statement.

ANALVARY

Thanks to Chuck Hopf, Dean Witter Reynolds, USA.

Change 05.6 PDB.JOBS observation built from job which had no job

Oct 22, 1986 termination record (happens if SPINCNT=0) caused minor

ANALMPL anomaly in program. Now, INBITS tested for job term.

Thanks to Chuck Hopf, Dean Witter Reynolds, USA.

Change 05.5 If ANALDSET was used with the PDB.STEPS data (rather

Oct 22, 1986 than using the SMF file and TYPE30_4 as step input),

ANALDSET a syntax error occurred due to a misplaced semicolon.

Thanks to Chuck Hopf, Dean Witter Reynolds, USA.

Change 05.4 Variables TAPE3420 and TAPE3480 are now created and

Oct 22, 1986 count the number of 3420 and 3480 tape devices which

VMAC30 were allocated by each step. TAPEDRVS is still the

VMAC434 total number of tape drives allocated and is the sum

VMACEXC2 of TAPE3420 plus TAPE3480. ANALTAPE modified to report

ANALTAPE on total drives as well as 3420 and 3480 drives.Also,

a better estimate of drives waiting alocation is made.

a. Add TAPE3420 and TAPE3480 to _VAR30 macro KEEP= for

TYPE30_4, TYPE30_5, and TYPE30_V data sets.

b. In VMACEXC2, replace line 166:

IF TEMPNEW THEN TAPEDRVS=SUM(TAPEDRVS,1);

with these five lines:

IF TEMPNEW THEN DO;

TAPEDRVS=SUM(TAPEDRVS,1);

IF DEVTYPE = 80X THEN TAPE3480=SUM(TAPE3480,1);

ELSE TAPE3420=SUM(TAPE3420,1);

END;

Thanks to Martha Hall, Metropolitan Life, USA.

Thanks to Ken Dice, Community Mutual Insurance, USA.

for ANALTAPE ideas.

Change 05.3 Variable RSPTMEDF, the RTM "Response Time Definition"

Oct 21, 1986 was undocumented and originally read in a PIB1. field.

VMAC39 It turns out that it is a Character byte, containing

FORMATS F - First Character Received, K - Unlock Keyboard, or

C - Change of Direction and thus describes which of

these definitions is being used in this 3274 RTM.

RSPTMEDF was change to $CHAR1. with new MXG format

$MG039DF supplied to decode the one-byte characters.

Thanks to Jim Yoney, Allegheny Power, USA.
Change 05.2 TSO/MON Command counts were incorrect in MXG, because

Oct 21, 1986 MXG counted all entries in the FCB "Full Command"

VMACTSOM segments as commands. Actually, the FCBs contain Full

Commands, Dialog Manager Programs, Dialog Manager

Panels (and in TSO/MON Version 5, CLIST executions

will also be counted in FCB segments). Only Full

Commands and Dialog Manager Programs should be counted

as commands, and TSO/MON has already stuffed the count

of these two in a CCB bucket with comand type of null

(i.e., hex 00). Remove line 3910:

NRTRANS=NRTRANS+NRTIMES inside DO _I_=1 to NRFCBS

to correctly count commands as TSO/MON does.

Thanks to Chris Moyle, SCP- working for MOBIL, AUSTRALIA.
Change 05.1 IBM re-numbered CPUID in the 3090-400. Instead of 0-3

Oct 21, 1986 they are 1-4. With this change, the CPU 4 individual

VMAC7072 measures will be stored in CPU0 variable names on the

3090-400.

CHANGE both occurrences of

IF CPUID=0 THEN

to read

IF CPUID=0 OR CPUID=4 THEN

This change was enhanced with Change 5.33.

Thanks to Lucy Fukishima, State of California Health & Welfare, USA.

LASTCHANGE: Version 5
=========================member=CHANGE04================================
/* COPYRIGHT (C) 1986 BY MERRILL CONSULTANTS DALLAS TEXAS */
MXG Software Status as of September 26, 1986.
This is the production release of MXG VERSION 4.4
Major enhancements in Version 4 are listed below. Read this

entire member to determine its impact. New changes have been

added, and some Change descriptions have been reworded since part

of these new changes were printed in MXG NEWSLETTER EIGHT in May.

Major items in Version 4:
1.New member INSTALL contains installation instructions for MXG

under either MVS or CMS versions of SAS. The MXG Exit Facility

is also documented therein. This member MUST be read to ensure

MXG installs without error.

2.VM Monitor processing to create VMONaaaa data sets directly with

SAS with VM MAP variable names. See change 4.85, below.

3.Support for RMF 3.4 and MVS 2.1.7 and Vector Processor data

introduced with 3090 processors. SMF manual TNL's thru GN28-1061

are included in MXG Version 4. New TYPE78CU and TYPE78IO data

sets and new data in TYPE78CF result from 3090 I/O architecture

4.IMSLOG CODE enhancement (TYPEIMS) puts log trace records together

for IMS transaction response and resource measurement.

5.CICS 1.6.1 and CICS 1.7 are now supported in VMAC110 code. New

program UTILCICS may have to be run to determine which of the

impacting PTFs are installed so that you can update IMACPTF.

6.Support for Landmark's The Monitor for CICS records, including a

INFILE exit to process compressed format data. See Change 4.94.

7.Support for Cullinet's IDMS Performance Monitor records.

8 ASM source for an exit to IDMS which will write SMF records that

contain IDMS log data, and SAS code to process those MXG-created

IDMS-LOG-SMF records. SAS code to process the IDMS DCLOG directly.

9.Support for DOS POWER Version 2 account records. Impacting change

in V2 to DOSJOBS data, and five new DOS data sets now exits.

10.Support for DISOSS Version 3 Release 3 account records.

10.New member IMACPTF. Presently used only for CICS 1.6 records, this

new maintenance feature allows MXG code to be enabled to support

those (dumb) PTFs which ambiguously change data format.

11.DB2 processing revised to capture all data segments in Release 2.

12.Revised ANALTAPE routine now with accurate counts.

13.MODEL204 accounting information now defaults ON (EXM24ACT).

14.NODUP option implemented in BUILDPDB/3 to remove duplicate SMF

input data from PDB data sets. (Full implementation for all data

sets requires SAS Version 5.15. See Change 4.38-39).

15.Too much more to list here. Read all of these changes! MXG now

creates 168 MXG SAS Data Sets (Version 1 had 78) with its 466

source library members (Version 1 had 189) processing data from

24 products for MVS, VM and DOS data centers.
/* COPYRIGHT (C) 1986 BY MERRILL CONSULTANTS DALLAS TEXAS */

What was not quite finished when development was frozen for 4.4?

1. CMS still has some known data errors in some fields. As far as

we know, all discrepancies are documented in the comments at

the beginning of VMACVMON. There is very little CMS reporting,

and essentially no guidance as to which variables in which data

sets are important. We have only validated the PERFORM, USER,

and DASTAP classes, though others are coded, they may not work

(yet). A very small number of users have tested non-HPO systems

and UP systems - our primary validation was 03.34.0314 HPO 3.4

on an MP. None of the planned EXECs for testing were completed

yet. Because of the newness of MXG support for VM Monitor data,

users should validate the data carefully, and share with us any

discrepancies. We will send MXG NEWSLETTERS if necessary to

communicate major CMS enhancements/discoveries over the next few

months. Because the vast majority of the data is valid, we felt

you would prefer to have this now rather than later. Please help

us with your comments and expertise.

2. Documentation! MXG Version 4 will be eventually enhanced by the

printed MXG SUPPLEMENT (a 3-ring document bound like manuals

for IBM-PCs). The SUPPLEMENT will be indexed to the same chapters

as the MXG BOOK, and will provide new sections of Chapter FORTY

for new and updated MXG data sets, as well as updates to other

chapters from all past MXG newsletters. A significant poriton of

the MXG SUPPLEMENT will deal specifically with the analysis of

VM Monitor data.

Merrill Consultants will automatically send one copy of the MXG

SUPPLEMENT to all supported customers when it is completed, and

future MXG Newsletters will be sized to fit therein.
When completed, additional copies of the MXG SUPPLEMENT will be

available for purchase (at a mere pittance) from the Publications

Department of SAS Institute. Please don't hassle SAS about the

publication date - they are not the hold up. Call us in Dallas

if you must, but don't expect it before year end. (Which year?)

COMPATIBILITY EXPOSURES BETWEEN MXG VERSIONS 3 and 4:

1. EXTY72 and EXTY74 exit members were changed. If they exist in your

USERID.SOURCLIB, your changes will override MXG changes. Thus only if

you have these members in your USERID. library, must you re-fit your

exit code. (The change externalized control of observations which had

previously buried inline in the MXG VMAC7072 member).
2. TYPE72 data set may contain more observations with Version 4.

Previous versions only output if performance group received SERVICE,

but RPGNs can be defined to only count transactions; MXG missed those.

Version 4 now outputs TYPE72 observations if either SERVICE or TRANS

are non-zero. This logic is now in EXTY72, if you wish to override.
3. Version 3 and RMF 3.3 cause the first TYPE70 observation from each

day's processing to have a missing value for the variable SU_SEC. This

is because that value is now in the type 72 record (at long last) for

RMF 3.3+, but the type 70 record (from which we used to have to get

the number by table look up) always precedes its type 72. Should only

affect report programs which depend on non-missing SU_SEC in every

observation of TYPE70. You'll have to get SU_SEC from TYPE72 in the

future, since that's where it should have been all along.

4. ROSCOE users must note Change 4.76.
Please call if you experience any difficulties or problems. Barry
This member describes the changes made to the MXG Software since

Version 3 was shipped. Changes to variables and data sets between

Versions 3 and 4 are documented in member DOCVER04. The total list

of the contents of all data sets, new and old, created now by MXG

Version 4 is in member DOCVER. VMON data sets in DOCVER with numeric

suffixes are not finalized, and may change in name and contents.

Always look for additional documentation in comments at the beginning

of each member. Note that the members affected by each change, below,

are listed under the date of change. Read those members too.
Additional documentation is available in past and future issues of

the MXG Newsletter, sent by Merrill Consultants to their Supported

Customers. Newsletter EIGHT was published May 31, 1986.
Note these SAS Zaps which you may need to install:
Z5152456 - 0C4, 30A when over 20 variables on PLOT statement

Z5152254 - Protected data set errors after PROC DATASETS

Z5162525 - ABEND 999 with ERRORABEND when no variables in data

set for PROC CHART, PLOT, GPLOT and maybe others.

The following pages provide the sequential log of changes which have

been installed already in this MXG library. The most recent changes

are listed first.
NEXTCHANGE: Version 4
=====Changes thru 4.111 as of September 26, 1986======================
Change 04.111 A new IBM error in CICS CMF records has been uncovered

Sep 26, 1986 and circumvented. Problem only occurs after other PTFs

VMAC110 (PP43887 and/or AP40463) are installed. MCTSSDRL (the

data record length) is 2-8 bytes shorter than the real

data to be created. IBM truncates the data segment to

the (incorrect) length specified by MCTSSDRL. An APAR

will be issued next week. MXG circumvents by comparing

the expected length (EXPLEN, from IMACPTF status) with

MCTSSDRL. However, two fields SECOUCHR and PROGSTOR

are overlaid by next segment, and will be missing in

CICSTRAN data set when error is circumvented. Another

error, hit when over 100 bytes of user data is added

was found in MXG and was fixed.

Thanks to Thom Conley, Gold Circle Stores, USA.

Thanks to Kirby Kern, Commercial Federal Omaha, USA.

Change 04.110 SAS 5.15 ABENDs with 0C4 or 30-A when more than 20

Sep 24, 1986 variables are specified on a PLOT statement. While SAS

ANALDALY zap Z5152456 fixes the problem, the offending PLOT

statements were split in MXG to circumvent problem.

Thanks to Chuck Hopf, Computer Language Research (FASTAX), USA.

Thanks to Drew Pierson, PRC, USA.
Change 04.109 The JES 3 Accounting information in the type 26 purge

Sep 24, 1986 record was not handled correctly, occasionally causing

VMAC26J3 a STOPOVER ABEND. Now, ACCOUNT1 is parsed directly

from the 42-character field. The STOPOVER dump Dave

sent showed an undocumented field, OUTDEVCE, has been

added to the Print section added with MVS 2.1.7.

Thanks to Dave Harrison, Celenese, USA.
Change 04.108 VM Monitor validation completed for MXG Ver 4.4. See

Sep 21, 1986 details in member VMACVMON. Almost all variables are

VMACVMON exact when compared with detail VM MAP listing with

TYPEVMON VM 3.4 HPO. Test sites have validated with non-HPO

ANALVMON VM Version 4. HPO 4.2 has been coded, but the new

variables have not been verified. Highly usable, but

this gem is only faceted - polish will come in next

Version of MXG.

Change 04.107 ESP site input has further validated and enhanced the

Sep 21, 1986 IMS Log processing. Code now conforms to MXG Exit

TYPEIMS structure. RESPONSE time & components (when possible)

are calculated. John enhanced the data kept and also

captured local data from the RACF segments.

Thanks to John Loo, Fluor Corporation, USA.

Change 04.106 More IBM CICS errors. With PP43887 on CICS 1.6.1, the

Sep 19, 1986 record length MCTSSDRL is not always correct in the

VMAC110 CMF type 110 record. Problem at IBM Level 2. Has

caused specious 'missing TASKNR message'.

Transaction JC.. sometimes has nulls for TASKNR which

caused "missing TASKNR" message. Fix by message only

IF TASKNR=. AND TRANNAME NE 'JC..' THEN DO;

Thanks to Tom Koelle, Citicorp Information Systems Research, USA.

Thanks to Joseph J. Faska, Chemical Bank, USA.

Thanks to Bill Cohen, Drexel, Burnaham and Lambert, USA.

Change 04.105 Validation of MXG DB2 data with DB2PRT caused fixes.

Sep 17, 1986 QTMAXDS=DIF(QTMAXDS) was deleted. MXG finds non-zero

ANALDB2 for QBnCRIO fields, which are zero on DB2PRT report.

QBnTCBA (n=1 to 4) disagree with DB2PRT at one site,.

but seem ok at another. Validate and call if you can.

SYSTEM added to SORT for DB2STAT1 and DB2STAT2 since

DB2 subsystem id (QWHSSSID) is not unique for SORT.

Thanks to Martha Hall, Metropolitan Life Insurance, USA.

Thanks to Scott Peterson, Southern California Edison, USA.
Change 04.104 VM Monitor cleanups

Sep 19, 1986 Non-HPO system: MN003CIE thru CSE do not exist. Fixed.

VMACVMON Used _HPO to test for class 0 code 3 to OUTPUT PERF &

VMONINST data sets after 0/03 record for non-HPO, but

OUTPUT them after class 0 code 4 for HPO systems.

Non-AP system does not write class 0 code 1 record.

Moved calculation of many variables to before OUTPUT.

Thanks to Wayne Kidd, SYNTEX, USA.

Change 04.103 MXG Lock manager data variables QTXA.... in DB2ACCT

Sep 16, 1986 have never been valid due to an MXG coding error. The

VMACDB2 QTXA.... variables in DB2STAT1 have been valid, tho

the IBM documentation is misleading. Now, QTXANPL

no longer exists in DB2STAT1 (it is only in ACCT) and

pointers to the six QTXA.... variables in DB2ACCT are

now read in from the correct offset. This fix was not

on 4.3 BETA tape.

Thanks to Sam Catalo, IBM Level 2, USA.
-----Changes thru 4.102 as of September 12, 1986 was the 4.3 BETA-----
Change 04.102 DISOSS Version 3 Release 3 Accounting record is now

Sep 12, 1986 supported. It provides more data than the perceeding

TYPEDISO record. See comments in TYPEDISO and DOCVER03.

FORMATS
Change 04.101 Turnaround example code now calculates in minutes to