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

converted into bytes and formatted with MGBYTES format,

so installed/used values can be compared easily, and the

labels indicate "storage" rather than "frames".

Many new variables are now created in ZRBASI and ZRBGEI.
This support is preliminary, in that it has not had any

extensive usage by real users, although I have had real

data to look at. Some of the fields may be accumulated,

and it may be necessary to add a post-processing step to

de-accumulate, but I need feedback from active users to

add the polish to this (long-overdue) enhancement. I did

not compare my output with snapshots of RMF III screens,

so please verify my calculations. The next interation of

ASMRMFV will provide record selection and will print a

summary report of records found/written. Originally,

ASMRMFV was named ASMMON3.

Thanks to Don Friesen, BC Systems, CANADA.

Thanks to Danal Estes, Logical Resources, USA.

Thanks to Lawrence Jermyn, Fidelity Systems, USA.

Change 12.258 JCLTEST6 may fail with LIBRARY MONITASK UNASSIGNED in

JCLTEST6 step TESTIBM2 or BUILDPDB, if you have changed member

Jan 27, 1995 IMACMONI to specify MACRO _LMONTSK MONITASK.MONITASK %.

JCLTEST6 invokes ASUMCICS in TESTIBM2 in member TYPE110,

and step BUILDPDB also includes ASUMCICS, and if you have

added the MONITASK ddname in IMACMONI, then both steps

will require that ddname. Although obscure, I have put

//MONITASK DD in both steps in the test job stream.

Thanks to Angela Mulcahy, UJB Financial Corp, USA.
Change 12.257 Support for TCP/IP Version 3.1 requires MXG 12.07 or

VMACTCP later to be compatible, but sites with MXG 12.01-12.06

Jan 27, 1995 can process Version 3.1 records by changing the LENGTH

Feb 3, 1995 tests (added by MXG Change 12.041)

from 200 to 204 for TCPEVENT='FTPSERVER'

from 86 to 90 for TCPEVENT='TELSERVER'.

TCP/IP Version 3.1 added FTPLCLUS (local user ID) and

FTPLCLPN/FTPRMTPN (local/remote port) in dataset

TYPETCPF (TCPSERVER), and added TELLCLPN/TELRMTPN in

dataset TYPETCPT (TELSERVER). (FTPLCLUS does exists in

records written by both TCP/IP Versions 2.2.1 and 3.1).

Thanks to Wanda Prather, The Johns Hopkins University APL, USA.

Change 12.256 Variable TRANSACT in Landmark is an eight-byte variable,

TYPETMON but only the first four bytes are used for transaction

TYPEMON8 name; the final four bytes are hex nulls ('00000000'X).

Jan 26, 1995 If you test for TRANSACT='ABCD', the test will always

fail, because SAS pads the literal 'ABCD' with blanks,

and ABCD-blanks is not equal to ABCD-hex-zeroes. You can

circumvent this effect by using the colon-modifier in the

equality: IF TRANSACT=:'ABCD' THEN ... which test true

for all values of TRANSACT that start-with ABCD. However,

I have decided to correct this problem at the source, and

have inserted the following TRANSLATE after each input in

both Landmark members:

TRANSACT=TRANSLATE(TRANSACT,' ','00'x);

Thanks to Tim Carne, Nottinghamshire County Council, ENGLAND.

Change 12.255 TYPE78CF variable PCTDIRPT is mislabeled as percent delay

VMAC78 and was miscalculated. The line

Jan 25, 1995 IF NRCMPTSM GT 0 THEN PCTDIRPT=100*PCTDIRPT/NRCMPTSM;

was replaced with

IF (CHPIDTKN+PCTDIRPT) GT 0 THEN

PCTDIRPT=100*PCTDIRPT/(CHPIDTKN+PCTDIRPT);

with new LABEL of 'PERCENT WHEN*DIRECTOR PORT*WAS BUSY'

Note: See Change 15.061, which revised this change.

Thanks to Thomas Heitlinger, FIDUCIA IT AG, GERMANY.
Change 12.254 -BatchPipes/MVS variables INTBTIME/INTETIME are in the

FORMATS SMFSTAMP8 format rather than TODSTAMP8 (obvious, when

VMAC91 you have test data records!). Find their INPUT after

Jan 25, 1995 IF SMF91PRL GE 60 THEN DO; and use SMFSTAMP instead of

Feb 9, 1995 TODSTAMP8 for INTBTIME and INTETIME (SMF91IST/SMF91IET).

Note: Do not change the first INPUT of only INTETIME, as

it (SMF91INT) really is in TODSTAMP format!

The logic for INTBTIME/INTETIME/DURATM is now INPUT only

for the interval subtypes 2 and 12 (their un-cleared, non

zero values were confusing). INTETIME is set to SMFTIME

and INTBTIME/DURATM will be missing in non-intervals.

-The lines with +4 after the INPUT of SMF91IWB and after

the INPUT of SMF91OWB must be deleted to correct the

values of SMF91IEC/SMF91OEC/AFSTTIME/ALSTTIME.

-Variables SMF91SST and SMF91STR are now decoded by format

for subsystem status (ACTIVE/STARTING/STOPPING) and trace

activity (ERROR/ FUNCTION/FLOW).

-APAR PN66074 corrects the value of SMF91PLR, the Pipe's

LRECL, which was 4 too large before the APAR.

Thanks to Lawrence Jermyn, Fidelity Systems Company, USA.

Change 12.253 MXG 12.04-12.06 only. ACF2 SMF record still gets an

VMACACF2 INPUT STATEMENT EXCEEDED RECORD LENGTH error results

Jan 24, 1995 because the code added by Change 12.072:

IF ACVMFIDC GT 0 THEN DO _I_=1 TO ACVMFIDC;

must be IF ACVMFIDC GT 0 THEN DO _I_=3 TO ACVMFIDC;

(the first two authids have already been input!).

Thanks to Peter Arcoudis, NRMA, AUSTRALIA.

Thanks to Dov Brosch, Solomon, USA.

Change 12.252 MVS/ESA 5.1, Goal Mode, TYPE72GO dataset variables from

EXTY72DL the Work/Resource Manager State Samples (all start with

IMAC7072 R723R... or R72XR...) were moved from TYPE72GO into new

VMAC7072 dataset TYPE72DL (for Delay, since these state samples

BUILDPDB report workload delays), because there can be more than

WEEKBLD one pair of Begin-to-End/Execution statistics for these

MONTHBLD "Transaction Service Class" observations (R723TYPE=3).

Jan 21, 1995 For example, if the SRVCLASS is a logical service class

(i.e., for CICS transactions), and those transactions

also call on IMS, the TYPE72GO for the SRVCLASS would

have delay state sections for both the CICS region and

the IMS region delays, but the original implementation

only kept the first set of delay statistics.

Additionally, these variables only exist in R723TYPE=3

records, so creating the new dataset will save DASD by

reducing the size of TYPE72GO dataset.

And dataset TYPE72DL will only have observations if you

have Transaction Service Classes, R723TYPE=3 records.

Thanks to Don Deese, (CPExpert), Computer Management Sciences, USA.
Change 12.251 DB2 Transit Report failed with NOT SORTED because the

ANALDBTR pairing of the 015-016-017-018 IFCIDs in ANALDBTR added

ANALDB2R CUBTOKEN (Change 12.124), causing the S015S018,S016S018,

Jan 21, 1995 S017S018, or S018S018 data to be out of sequence with the

rest of the trace events. The fix in ANALDBTR was to add

PROC SORT DATA=SxxxS018 OUT=SxxxS018;

BY SYSTEM QWHSSSID QWHSACE QWHCCN QWHCCV BEGEVENT;

for each of the four SxxxS018 datasets at the end of the

_015PAIR macro. In addition, the design of the Transit

Report was changed to minimize DASD required. The report

previously built the TRANSIT dataset (747 variables!) and

then did a PROC PRINT; now, the report is produced with

PUT statements, and the TRANSIT dataset is now built with

only 6 variables (for a summary report of event counts).

Thanks to Dan Null, Caterpiller, USA.
Change 12.250 DB2 Locking Contention Report PMLOK02 in ANALDB2R was

ANALDBTR incorrect, and the pairing logic for 044-045-054 IFCIDs

XDB2LOCK in ANALDBTR was also incorrect. ANALDBTR has been fixed

Jan 20, 1995 while PMLOK02 is still being repaired. However, member

XDB2LOCK provides an analysis of DEADLOCK/TIMEOUT events

that may meet your needs, and can be easily tailored if

you need more details on these contention events.

Thanks to Charlie Royster, AHOLD, USA.

Change 12.249 The DEVCYL value (cylinders per device) for RAMAC devices

VMXGVTOF is different than for a native device. For native 3390-3

Jan 16, 1995 DEVCYL=3340, but for RAMAC 3390-3, DEVCYL=3339, which is

the number of cylinders that you can really use! So why

is the value different? Because native devices have 15

alternate tracks (one cylinder), which DEVCYL counts, but

RAMAC alternate track support is internal, and does not

count the extra cylinder. In VTOC processing, MXG has

always subtracted one from DEVCYL so that DEVCYL*DEVTRK

would give you the correct capacity, but for a RAMAC VTOC

DEVCYL ends up one cylinder too small!

(The impact of one cylinder of capacity would probably

be overlooked, but this site used exact values of DEVCYL

in a FORMAT to identify which model 3380/3390 was used,

and the off-by-one value fell thru their format!).

At present, there does not seem to be a way to tell that

the VTOC is from a native device versus a RAMAC device,

so I may have to resort to a table of expected sizes, but

I await IBM's response.

Thanks to Paul Polley, PolyGram Holding, Inc, USA.

Change 12.248 The //IMSSUM DD had RECFM=F,LRECL=132,BLKSIZE=132 when it

ASMIMSLG should have been RECFM=FB,LRECL=132,BLKSIZE=23364. This

Jan 16, 1995 can only help the elapsed run time! (Also, the LRECL=132

files with BLKSIZE=13200 were increased to BLKSIZE=23364,

for half track on 3380s.).

Thanks to Don Cleveland, Blue Cross and Blue Shield, USA.

Change 12.248A Cosmetic. The comments and examples for using IMACFILE

IMACFILE were revised, and the list of variables available to the

Jan 16, 1995 exit was corrected.

Thanks to Mr. Loewenthal, Zonussi Elettrodomestici, ITALY.

Change 12.247 Variable LSCDSTYP replaced a reserved field in NPMINSES

VMAC28 dataset in NPM 2.2, but was overlooked until now.

Jan 16, 1995

Thanks to John Steigerwald, Canadian General, CANADA.

Change 12.246 Matching CICSTRAN with DB2ACCT found occurrences when the

ANALDB2C DB2 plan continued to execute long after the CICS trans

Jan 12, 1995 had ended, and a trace of these transactions was needed,

so a new %MACRO DBUGDB2C will now print out the detail

events for a selected NETSNAME/UOWTIME value. Also, the

matchup logic now prints notes on the log on selected

mismatches (first 5, then every 100,000th) so you can

diagnose data problems (like one day's DB2 with another

day's CICS!). See comments in the member, and the DB2

Technical Note in Newsletter TWENTY-SEVEN.

Thanks to Ed Long, Fidelity Systems, USA.
Change 12.245 INVALID DATA FOR OPENDTE in type 14/15 records with nulls

VMAC1415 in that date field are corrected by adding ?? between

Jan 12, 1995 OPENDTE and PD4. and by wrapping the OPENTIME and OPENTM

calculations with IF OPENDTE GT 0 THEN DO; ... END;

I do not know why these records, clearly written after

APAR OW00484 was installed, have zero date, but only one

such record has been found.

Thanks to J. R. Bleeker, Newport News Shipbuilding, USA.

Change 12.244 Variable LCU or LCUID was FORMATted as HEX2., but sites

DOC now have more than 255 LCUs, so the variable was changed

Jan 11, 1995 from HEX2. to HEX4. in members TYPE8911,TYPE10,TYPE1415,

TYPE19,TYPE21,TYPE64,TYPE69,TYPE74,TYPE75,TYPE78,TYPE79.

Thanks to John Astle, National Australia Bank, AUSTRALIA.
Change 12.243 Support for ICEBERG's IXFP SRP Subsystem Performance SMF

VMACICE record PUT9404 changes the subtype 1 record (MXG dataset

Jan 11, 1995 ICEBRGSY) compatibly, adding variables FREBESCT/FREBESCP

(free collected back end space total/in prod partition).

Change 12.242 Support for DFSMSrmm records with DFSMS Release 1.2, and

TYPEEDGB new support for rmm's Control Backup file.

VMACEDGR To summarize MXG support for various rmm data sources:

VMACEDGS Member Infile Reads

Jan 11, 1995 TYPEEDGB LOGSMF Control Backup (EDGSKIP with BACKUP)

TYPEEDGR EDGHSKP Extract (EDGSKIP with RPTEXT)

TYPEEDGS SMF SMF Audit and Security (must update

member IMACEDGS for SMF record ID).

-VMACEDGR support for the DFSMSrmm Extract Record type S

causes INPUT STATEMENT EXCEEDED because the +38 Reserved

field following RSBINNO should have been +31. Also,

field RVSTSTAT does not exist, RVSTBIN and RVOBIN should

have been $EBCDIC6 instead of $EBCDIC8, and RVSTDATE

should have been $EBCDIC10 instead of $EBCDIC8.

This Extract Record is created by EDGHSKP with RPTEXT

option, but all of the dates are in character instead of

SAS dates, so this file may be less than useful.

-However, if EDGHSKP is run with the BACKUP option, that

file creates records that are almost like the SMF records

that are processed by VMACEDGS, except that there is no

SMF header, and there is no EDGS header. With minor

changes to VMACEDGS, new member TYPEEDGB will now read

the backup file (but note, you must use the //LOGSMF DD

to point to the backup file). This seems to be the best

data to use account for and analyze all your rmm tapes!

-The VMACEDGS Audit Record type D variables MDPDSN/MDNDSN

were always INPUT, but they may or may not exist; now the

INPUT is conditional. based on MDPDSNL/MDNDSNL.

Thanks to David Froberg, Software AG, USA.
Change 12.241 MVS/ESA Goal Mode type 72 subtype 3 does not measure the

VMAC7072 SYSTRNTM, the total system transaction time, from read in

Jan 10, 1995 to termination. Instead, variable AVGXETTM (Average

Execution Time), which is a subset of the AVGELPTM

(Average Elapsed time) duration is created. (SYSTRNTM

does exists in Compatibility Mode.)

Thanks to Don Deese, (CPExpert), Computer Management Sciences, USA
===Changes thru 12.240 were included in MXG 12.06 dated Jan 09, 1995===
Change 12.240 Cosmetic changes (LABELs for unlabeled variables, names

DOC in comments corrected, delete of temporary datasets that

Jan 8, 1995 should not have been kept, uncovered during QA runs) were

made to these 18 members:

ANALVMDY ASUMTALO INSTALL JCLINSTL JCLUXRE6 TESTIBM3

TYPEMON8 TYPETMON TYPE102A VMACRSPS VMACSIM VMACTIRS

VMACTLMS VMACVMON VMACXAM VMACZRB VMAC0 VMAC44
Change 12.239 Dan Kaberon has revised his analysis that merges the MVS

ADOCPATH TYPE73, TYPE74, and TYPE78CF data to analyze I/O path

ANALPATH statistics. An excellent documentation of the report is

ZNALPATH also provided by Dan in ADOCPATH, and (though unlikely to

Jan 8, 1995 ever be needed again), the original ANALPATH is ZNALPATH.

Thanks to Dan Kaberon, Hewitt Associates, USA.

Change 12.238 Almost cosmetic. There was no EXDB2STS exit member for

EXDB2STS dataset PDB.DB2STATS, but as there was also no statement

DIFFDB2 "%%INCLUDE SOURCLIB(EXDB2STS);" in DIFFDB2 (the OUTPUT

Jan 8, 1995 statement was implicit). Now there is an INCLUDE to

explicitly invoke the (newly created) exit member.
Change 12.237 MXG 12.04-12.05 only. Support for new subtype 5 Iceberg

EXICEDEL Deleted Data Space Release event (MXG dataset ICEBRGDE)

VMACICE was incorrectly output in dataset ICEBRGDR, and there was

Jan 8, 1995 no exit member EXICEDEL added to the Source Library,

because the "%%INCLUDE SOURCLIB(EXICEDRV);" following the

"MACRO _CICE05" definition in member VMACICE should be:

"%%INCLUDE SOURCLIB(EXICEDEL);/*_LICEDEL OUTPUTS ICEBRGDE*/
No one reported this error (which actually has minimal

impact since the subtype 5 is new and infrequent).

Instead, it was detected by a real fine addition to my QA

jobstream that was written by Freddie Arie. Using the

structure and naming conventions of MXG, Freddie reads

the MXG source library to look for exceptions and lack

symmetry between the IMACs, VMACs, and the exit members,

such as mismatched comments and datasets not output, and

it even verifies that any dataset name in a comment after

each OUTPUT statement is actually the dataset name that

is ultimately output! His powerful logic found my coding

error (I had block replicated code to add support for the

new subtype 5 record, but had failed to change the exit

name!).

Thanks to Freddie Arie, Lone Star Gas, TEXAS.
Change 12.236 DB2 report PMSQL01 may fail because QWHCTOKN was in the

ANALDB2R BY list for the report step, but was left out of the BY

Jan 8, 1995 list for the sort (and the failure is only apparent if

your data has non-blank QWHCTOKN!). Replace the line

NETSNAME QWHSLOCN BEGEVENT; with

NETSNAME QWHSTOKN QWHSLOCN BEGEVENT;

Change 12.235 Further cleanup of CICS reporting. Correct summarization

ANALCISH of "Storage Manager" CICSMD/CICSMT/CICSMDSA and "Journal"

Jan 5, 1995 CICJCR. "Last Reset time" (DSGLRT) was incorrect on the

report headers. Merged CICxxxxx datasets with CICDS, BY

statement changed to BY SYSTEM COLLTIME;, and VMXGSUM is

now used internally.

Dataset summarization status with this revision:
CICVT "VTAM STATISTICS"

Summarization value (times at RPL maximum) does

not match IBMs value, DETAIL values do match.

CICTCR "TERMINALS"

CICSMD-CICSMT-CICSMDSA "STORAGE MANAGER"

CICJCR "JOURNALS".

CICTCLR "TCLASS STATISTICS"

CICTM "TABLE MANAGER"

CICAUSS "AUTOINSTALLED TERMINALS"

CICAUTO "AUTOINSTALLED STATISTICS"

Summarization value (Times the peak was reached)

does not match IBMs value, DETAIL values do match.

CICDBUSS "DBCTL STATISTICS"

CICLDR "PROGRAMS"

Summarization value (Average fetch time) does not

match.

CICDS "DISPATCHER & TCB STATISTICS"

CICDTB "DYNAMIC TRANSACTION BACKOUT"

CICFCR "FILES DATA TABLE STATISTICS"

CICIRCB "BATCH GLOBAL STATISTICS"

CICLDG "LOADER STATISTICS"

Value in question: "programs loaded but not in use".

CICM "MONITORING STATISTICS"

CICDQG "TRANSIENT DATA GLOBAL"

CICDQR "TRANSIENT DATA RESOURCES"

CICLSRR "LSRPOOLs "

SHARED DATA BUFFERS O.K.

Had no index data or hiperspace data to verify values

CICLSRFR "LSRPOOL FILES"

CICST "STATISTICS DOMAIN STATISTICS"

CICTC "TASK ACCUMULATED SO FAR"

CICTSQ "TEMPORARY STORAGE"

CICCONSS "ISC/IRC ATTACH TIME STATISTICS"

average reuse time values do not match:

IBM Average reuse time tween entries: 00:00:08.3886

MXG Average reuse time tween entries: 0:00:00.296

and the MXG value was summed.
NEW CICS 4.1 REPORTS. None have been validated with 4.1

test data yet; this note will be updated when tested.

CICDLIG "DL/I GLOBAL"

CICFEPIC "FEPI CONNECTION"

CICFEPIP "FEPI POOL"

CICFEPIT "FEPI TARGET"

CICPAUTO "AUTOINSTALL PROGRAM"

CICPUSG "USER DOMAIN"

CICXMC "TRANSACTION MANAGER TCLASS"

CICXMG "TRANSACTION MANAGER GLOBAL"

CICXMR "TRANSACTION MANAGER TRANSACTIN"

Thanks to Mike Major, RSCA, USA.

Change 12.234 This is a documentation change only. No code was changed.

ASMTAPES The Tape Allocation Monitor portion of ASMTAPES still has

Jan 5, 1995 occasional incorrect allocation records (see below), but

ASMTAPES still is the replacement that should be used

in place of ASMTMNT, because the Tape Mount Monitor part

of ASMTAPES is better than ASMTMNT. Comments in ASMTMNT

explicitly tell you to use ASMTAPES instead of ASMTMNT!
Note that if you are running an old copy of ASMTMNT, and

if you suddenly find there are no tape mount records, it

probably means that your site installed the new HCD

architecture (which supports dynamic addition of new tape

devices without a SYSGEN). You will need to re-assemble

ASMTAPES and you will find your mount records will then

reappear. (Some sites assembled ASMTMNT four years ago,

before it was modified for the new HCD architecture!).

These problems are still not resolved in MXG 12.12:
-There are occasional, apparently specious, records with

PROGRAM='IEFIIC' or PROGRAM='STARTING', and typically

with a very short allocation duration that should be

deleted from your analysis. We are still investigating

why these incorrect program names and very short

allocation durations are created; statistically they seem

insignificant, but pragmatically we need to trap and

understand them. I though about deleting them for you,

but that runs against the grain, so I report them here

while the investigation is in progress. I suggest you

look at your data first, and then delete these

observations (and any with a duration less than ten

seconds, as these are insignificant events in the big

picture).

-Thanks to very fine research by Michael Enad of Dun &

Bradstreet, two additional problems in the Allocation

monitor are under development - we apparently miss two

types of deallocation events:

- When a tape volume is swapped from one drive to

another (either due to an error on the first device,

or due to operator command to swap the mount into or

out of a silo), we currently miss the deallocation of

the first device, and instead of writing two separate

allocation records, only one record is written when

the last device is freed, and that record still has

the device address of the first device, so if a

different job happens to allocate the first device

after the swap, that legitimate allocation record

appears to overlap the incorrect allocation record of

the job whose tape was swapped.

- When HSM allocates a device for Migration/Recall and

has multiple mounts, and then later the same day HSM

allocates that same device for Backup, two Allocation

records are written (both at end of the Backup), but

the deallocation time for the Migration/Recall event

has the end of Backup timestamp, causing these two

records to overlap.
Change 12.233 Continued enhancement of what will become VMXGSUM, but is

XMXGSUM currently in XMXGSUM for additional parallel testing.

Jan 5, 1995 Parsing now recognizes the syntax of all options on any

data statement. Output variables are created only if the

variable exists in the input data set (so you can now

drop variables from your DB2ACCT and still run ASUMDB2A

to summarize only your kept variables - previously, all

variables named in ASUMDB2A were created in the output,

even when they did not exist in your DB2ACCT dataset).

See discussion in the member.

Thanks to Chuck Hopf, Merrill Consultants, USA.
Change 12.232 Labels for FIXEDAV,CSLPFXAV,LPAFXAV,LSQAFXAV,PRVFXAV,

RMFINTRV SQAFXAV,FIXLOAV,PVTAFCAV now state '...*MAX OF AVERAGE'

Jan 5, 1995 instead of AVERAGE. In summarizing these measures of

real frames, taking the real average of all intervals

results with an un-representative number, because of the

slack intervals. By calculating the maximum value of

the averages for each interval, I think you get a much

more representative value, but I should have also made