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

Change 13.120 Support for Boole & Babbage's HiperCache 4.1.1 and 4.1.3,

VMACHIPR (INCOMPATIBLE, because data was inserted into the fixed

Jul 18, 1995 portion of the record and LSR segments were expanded).

The JCTJOBID,INITTIME and READTIME were added to the VSAM

record, and LSR subpool information added new variables

SMFnnHSB,HSR,HSW,HFR,HFW,MFH,MTH,FFH, and FTH for buffers

in hiperspace, and successful/failing reads/writes

and move-pages to/from hiperspace.

Thanks to Dave Myers, Time Warner Cable, USA.
Change 13.119 XMXGSUM update. This new enhanced replacement for VMXGSUM

XMXGSUM was ready to go, until I ran TESTANAL and discovered that

Jul 18, 1995 repetitive invocations of the new logic caused strange

errors (Unable to Load IMAGE SASQUIL, Unable to Open

File) due to memory problems. Increasing the REGION= to

48M eliminated the errors, but until I examine why there

is an increase, it is not safe to replace VMXGSUM yet.

But because you are wise enough to read this text, and

can be prepared to increase region size to save lots of

CPU, time, and DASD space, I strongly recommend you use

the new XMXGSUM (see comments therein) for summary of the

large datasets (CICSTRAN, DB2ACCT in particular).

Change 13.118 -DB2 Statistics Detail and Long Trace reports had EDM pool

ANALDB2R variables that were non-zero for never-happened, and some

ADOCDB2 totals were printed instead of averages. One logic error

Jul 18, 1995 (repeated variables in SUM= list) was not caught because

the test suite uses the new XMXGSUM replacement for

VMXGSUM, and one of the many enhancements in the XMXGSUM

replacement is that it weeds out any duplicates! Using

XMXGSUM will correct some of the report values, but not

all, so you will still need to install the new ANALDB2R.

-Audit reports show FORMAT $DB2OBID NOT FOUND, even though

the log shows they were created. I think the error is an

out of memory condition while loading this very large

format, but there is no clue on the log!

But the real culprit is the creation of the TEMPDBID and

TEMPOBID datasets that are used as CNTLIN= for PROC

FORMAT. The early logic to map DBID/OBIDs output one

value for each T102S105/T102S107 obs, (open), but a new

value is needed only if the mapping changed, so the logic

now outputs one value for each unique mapping. The

500,000 obs in TEMDBDID dropped to 2,000 and the memory

problem disappeared.

Increasing MEMSIZE and REGION may circumvent, but 72MB

was not enough with 500,000 obs in TEMPOBID dataset

that is used as CNTLIN= for PROC FORMAT.

-ADOCDB2 was updated with an IBM note on why SRB CPU time

in DB2ACCT is invalid for CICS attachment and for DDF.

-PMSQL01 trace was rewritten as a series of sorts and one

monster data step using SET and a BY statement to save

DASD space. Major reduction in DASD space with this

revisions (200 cyl now, 3000 cyl not enough before!).

Time selection is only at the report level for SQL01;

all data still must be read and processed (hence put on

temporary DASD). We are reviewing alternatives for the

next version, but this change makes SQL trace feasible

even with 500,000 trace events in only 200 cylinders of

WORK space (previous version failed to complete with an

entire 3000 cyl volume!). You can always use IMACFILE

to select SMFTIME of interest with PDB=SMF execution

until we have time to address this enhancement to SQL01.

Thanks to Phil Schwartz, Pershing, USA.

Thanks to Dean Berry, Farmland Industries, USA.

Thanks to ???, Schweizerische Bankgesellshaft, SWITZERLAND.

Thanks to Clark Jennings, Reynolds Metals Company, USA.
Change 13.117 Variables were not kept because they were misspelled in

VMACCADM the KEEP= list. Variables STFKUS18 STIVSR10 STIVSR16

Jul 15, 1995 STIVSR22 STIVSR28 and STIVSR34 replace their misspelling.

Thanks to Freddie Arie, Lone Star Gas, TEXAS.

Change 13.116 Landmark converted History Records with invalid data for

TYPEMON8 MRO segment cause "INVALID OFFSETS IN USER SEGMENTS".

Jul 15, 1995 The record is defective, and Landmark is investigating

their error, but MXG, by default, DELETEs the record so

there is no observation output. To circumvent, insert

MROCOL=0; MRONUM=0; after MROCOUNT=0; so that the MRO

logic will not be executed. This will cause zero obs in

the dataset MONIMRO, and variables MROTIME,MROCOUNT will

always be zero in dataset MONITASK, but dataset MONITASK

will contain valid observations. This circumvention will

NOT be installed in MXG software; it is printed here only

for user modification until Landmark fixes their error.

Thanks to Warren Tanaka, Farmers Insurance Group, USA.

Thanks to Bill Padilla, Farmers Insurance Group, USA.

Change 13.115 TYPE42 variables AVGCONMS AVGDISMS AVGPNDMS had format

ADOC74 TIME13.3, but those MS PER SSCH variables should be 5.1

VMAC42 (so that 12.1 milliseconds prints as 12.1 instead of

VMAC74 0:00:12.100!). If BUILDPDB includes type 42 records,

VMACVMXA the TYPE74 variables will also have wrong format. The

Jul 12, 1995 -VMAC42. Move AVGCONMS AVGDISMS AVGPNDMS from TIME13.3 to

5.1 format in VMAC42, labels were made consistent.

-VMAC74. Labels were made consistent.

-ADOC74. Variable AVGENQMS noted as always missing in XA.

-VMACVMXA. Labels were made consistent.

-VMAC74. Also, variables AVGPNDIR PCTPNCHA PCTPNDIR were

added to 5.1 format.

Thanks to Tim Carne, Nottinghamshire County Council, ENGLAND.
Change 13.114 Variable TPXSNAME (Session Name) was always blank. Change

VMACTPX IF SKIP GT 8 THEN DO; to IF SKIP GE 8 THEN DO;

Jul 12, 1995 TPXSNAME only exists in dataset TPXAPLON.

Thanks to Jim Tintle, Mack Trucks, Inc, USA.

Change 13.113 MXG 13.01-13.02B only. CICS Shutdown report may cause

ANALCISH NOTSORTED error.

Jul 12, 1995 -Move %REPORTS statement to before %IF &REQ_REP GT 0 ....

-Find the %VMXGSUM invocation with INDATA=CICTM. Change

NOSORT=YES to NOSORT=NO in that invocation. Remove the

statements PROC SORT DATA=&MXGDS OUT=&MXGDS %VMXGFOR;

BY &BYLIST; %LET BYLIST=SYSTEM APPLID;

that immediately precedeed that %VMXGSUM invocation.

Thanks to Chris Weston, SAS Institute, ENGLAND.
Change 13.112 ACF2 subtype "L" (dataset ACF2JR) logic was redesigned.

VMACACF2 The old logic created one before ARE obs and two after

Jul 12, 1995 ARE obs, even when there were multiple ARE segments, and

the LID was decoded but not output. The redesign creates

new dataset ACF2JRL to contain the LID fields (and the

identity of the changer), one obs in ACF2JRL for each "L"

record, and dataset ACF2JR now has one obs for each ARE

entry (either before or after) with the ARE field name

and field value and the identity of the changer.

While I expected a Before and an After ARE for each

field that was changed, that's not exactly what ACF2

writes out. For fields that are bit switches, those

fields only show up in an ARE when their value is 1.

Turning a bit switch on has no Before ARE but has an

After ARE. Switching a bit switch off has a Before ARE

with no After ARE entry, so Before/After AREs cannot

be paired, but ACF2JR dataset has all the information.

Thanks to Normand Poitras, Banque Nationale du Canada, CANADA.

========MXG Version 13.02B, dated Jul 6, 1995, thru 13.111======
Change 13.111 Several variables were not kept due to spelling errors,

IMACHURN macro _KHRN527 was misspelled in IMACHURN, and datetime

VMACHURN variables were not stored as LENGTH 8.

Jul 6, 1995

Change 13.110 MXG Position Paper on support for the Year2000. MXG now

YEAR2000 supports dates after 1999, but many data records from

Jul 6, 1995 many products do not. This paper identifies which fields

in which record from which vendor do not have field width

to support Year2000, and which fields must be validated

by vendors that, while wide enough, do in fact support

the Year2000. This member will be updated as vendors

make changes and verify their date formats.

Change 13.109 Only comments were changed, but a new example of how to

IMACFILE select CICS type 110 records by APPLID or SUBTYPE was

Jul 6, 1995 coded and added as a comment block.
Change 13.108 HSM dataset DGN only processed first dump copy when more

VMXGHSM than one copies were created at the same time; copies

Jul 6, 1995 will have same start date/time and source volume in the

base, but dump copies will have different dump class name

and tape volser, and may have different output units and

device types. Additionally, this correction adds vars

hh0 mm0 ss0 and tt0 to the keep= list (so that the MCDS

key can be reconstructed).

Thanks to Wanda Prather, Johns Hopkins Applied Physics Lab, USA.
Change 13.107 Temporary datasets used in these members were kept when

ANALCNCR they should have been deleted at the end of processing.

ANALINIT This is mostly cosmetic (some of these datasets ended up

ANALTAPE being documented in member DOCVER!) but it is MXG's

TRNDTALO intention to not leave datasets in the //WORK library.

Jul 5, 1995

Change 13.106 DB2 Reporting was still wrong in 13.02A. The Statistics

ANALDB2R Reports produced replicated pages and wrong totals, due

Jul 5, 1995 to a mislocated block of code (introduced in MXG 12.12).

Also, MXG 13.01 introduced a FORMAT NOT FOUND error for

DB2DBID or DB2OBID format in Audit Reports if there were

no IFCID=105/107 observations found; the Audit report was

enhanced to print the tablespace name instead of the hex

OBID/DBID value, but the Audit report logic did not tell

you that you needed the 105/107s to create the format!

Now, we will use 105/107s if found, but will not fail if

they do not exist. (However, our Format building logic

may require REGION=48M and MEMSIZE=48M in your CONFIG

member in this version, due to duplicate entries - we

know how to fix but not in time for MXG 13.02B.)

Finally, the SQL trace report can take a lot of DASD

space, and we are re-designing that report for better

execution performance in the next version, and we will

update the instructions in comments in ANALDB2R for the

IFCIDs needed/used in various reports.

Thanks to Dean Berry, Farmland Industries, USA

Thanks to Neil Ervin, Huntington Service Company, USA.
Change 13.105 DCOLLECT with SMS 1.2 fails with INPUT STATEMENT EXCEEDED

VMACDCOL error, and prints a hex dump of a 260-byte record, if you

Jul 5, 1995 still have LRECL=264 specified in the JCL for DCOLLECT.

With SMS 1.2, the LRECL=644 must be specified for the

output of DCOLLECT (see Change 12.316, only in 12.12A),

but I did not cross-reference that change to this error.

Change 13.104 Variable GMTOFFTM was missing if LENPROD=60 (but this had

VMAC7072 no effect, as GMTOFFTM is not actually used in MXG, it is

Jun 29, 1995 only a kept variable in TYPE70). GMTOFFTM=GMTOFF70; was

inserted after GMTOFF70=GMTOFF70/4096;

Thanks to Don Friezen, B.C. Systems, CANADA.
Change 13.103 Support for 4-digit UCB addresses in Cache RMF Reporter.

VMACACHE Variables DEVN and DEVPLX need to be formatted HEX4..

Jun 29, 1995 Originally, I changed the input of @9 UNITADDR $EBCDIC3.

Jul 12, 1995 to @8 UNITADDR $EBCDIC4., hoping that IBM had the wisdom

to use that extra preceding byte for the UNITADDR, but

no joy, the byte @8 is still a reserved, hex 00 byte.

So UNITADDR in CACHE90 dataset is only 3 characters, and

you must use DEVN variable for the device number (the new

name for unit address). This note revised Jul 12, 1995.

Thanks to Rick Balint, Seattle First, USA.

Change 13.102 Support for AUTOMATE had some incorrect field lengths in

VMACAUTO the initial coding. Input of SMFA_SSN was changed from

Jun 29, 1995 $EBCDIC3. to $EBCDIC4., and SMFA_NAM from $EBCDIC3. to

$EBCDIC8. Also, SMFA_ST was added to DATETIME21.2 format

and SMFA_SSN was added to the KEEP= for all datasets.

Thanks to Rik van Roy, Procter & Gamble ETC, BELGIUM.

========MXG Version 13.02A, dated Jun 28, 1995, thru 13.101======
Change 13.101 VM CPU ID values of 'A01234'X were not supported by the

TYPEVM &PK.3. informat used in MXG, so the input was changed to

Jun 26, 1995 &PIB.3., and formatted HEX6. to support all hex valued

in the CPU ID field.

Thanks to Mr. Koeller, Sparkassen-Datendienst AG & Co. KG, GERMANY.
Change 13.100 MXG inputs DOS dates using MMDDYY8. format, because that

TYPEDOS is the way IBM writes them in the US. However, if your

Jun 26, 1995 site uses European format dates under your DOS, you will

have to change the three MMDDYY8's to DDMMYY8 because

there is no way to detect from the record.

Thanks to Mr. Koeller, Sparkassen-Datendienst AG & Co. KG, GERMANY.

Change 13.099 All nine occurrences of DATE &PD.4. were changed to

VMACFTP DATE ?? &PD.4. to suppress error message and hex dump

Jun 26, 1995 if invalid (or null) dates are found in the record.

Thanks to Mr. Koeller, Sparkassen-Datendienst AG & Co. KG, GERMANY.

Change 13.098 DB2 Accounting fields with repeated commas create SMF 101

VMACDB2 records with repeated 'FF'x values, causing MXG parsing

Jun 26, 1995 logic to fail, generating error messages on the log.

Replace all nine occurrences of IF LOC=0 THEN ...

with IF LOC LE 1 THEN ...

Thanks to Stephen Bell, Westfalisch Lippischen Sparkassen, GERMANY.

Change 13.097 Macro code which stripped blanks out of INDATA operand

XMXGSUM was removed; it was no longer needed (the XMXGSUM logic

Jun 26, 1995 now parses individual words of the INDATA string) and it

was found to consume significant CPU time.

Thanks to Mitch McKenna, SAS Institute Europe, GERMANY.
Change 13.096 DB2 Statistics Reports were corrupted in MXG 12.12 when

ANALDB2R support for DB2 3.1 was added (a block of code was mis-

Jun 26, 1995 located). See Change 13.106.

Thanks to Mitch McKenna, SAS Institute Europe, GERMANY.

=======Spoke at Greater Boston CMG meeting in Waltham, MA, 21Jun95======
Change 13.095 TYPE72 variable DOMAIN (SMF72CDN) has a value of '5c5c'x

VMAC7072 for Report Performance Group observations, because there

Jun 19, 1995 is no domain for RPGN's - it is the Control Performance

Group that measures resource consumption by Domain.

(My transaction in RPGN=99 could span several Domains as

work changes Periods in its Control Performance Group)

Thanks to Maurice Clark, ISSC Australia, AUSTRALIA.

Thanks to Chris Nibbs, ISSC Australia, AUSTRALIA.

========Initial MXG Version 13.02, dated Jun 19, 1995, thru 13.095======
Change 13.095A Variable DEVNR in type 42 datasets was not formatted,

VMAC42 causing device 0181x to print as 385! Add to FORMAT:

Jun 19, 1995 DEVNR HEX4.

Thanks to Mark Robbins, W. H. Smiths Limited, ENGLAND.

Change 13.094 Variable LUNAME was skipped in the 'E3'x record causing

VMACNAF other variables in dataset NAFLOGOF to be trashed. After

Jun 19, 1995 RTYPE='E3'X test, after NAFRLN $CHAR4. insert

LUNAME $EBCDIC8. (and add to the KEEP= list for dataset

NAFLOGOF).

Thanks to Mark Hayden, General Accident, SCOTLAND.

Change 13.093 Support for DFSORT Release 13 (INCOMPATIBLE) added many

EXTY16IN new variables to TYPE16, and three new datasets are now

EXTY16SO created with details on each file used:

EXTY16OU TYPE16IN - Input Data Sets (including concatenations)

IMAC16 TYPE16SO - SORTOUT Data Sets

VMAC16 TYPE16OU - OUTFIL Data Sets.

Jun 17, 1995 Note that DFSORT Release 13 comes with a boost that is

Nov 15, 1995 specifically designed to improve sorting of SAS datasets

(and at no additional cost!) - see the IBM announcement!
Change 13.092 Support for OPC Release 3.0 (INCOMPATIBLE). IBM added 12

IMACOPC new bytes for TRLLENGT in the header, inserted 16 bytes

VMACOPC for new field MT0CAL for dataset OPC24_1, and added new

Jun 17, 1995 MTDTYPE=15 for Step Restart (new MXG dataset OPC24D_J).

To process Release 3 records, you can make these changes

until you get the new MXG version:

-Insert after the @; that is after the INPUT of TRLFILL:

IF _OPCVER GE 3 THEN DO;

INPUT +12 @;

OFFSMF=OFFSMF+12;

END;

-Insert after the @; that is after the INPUT of MT0OTX:

IF _OPCVER GE 3 THEN INPUT +16 @;

-Update your IMACOPC to set macro _OPCVER to 3. I chose to

leave the default Release in IMACOPC unchanged at 2.

This code has been tested against only some OPC records;

there may be additional changes if other subtypes were

changed, and this text will change if necessary.

Thanks to Mr. Riccardi, AGIP Petroleum S.p.A., ITALY
Change 13.091 Support for Legent's AUTOMATE User SMF record creates 11

EXAUTOxx new datasets. Of course, now that the work is done, I

IMACAUTO discover it is Legent's intention to stabilize AUTOMATE

TYPEAUTO and replace it with OPS-MVS! In any event, there is a

VMACAUTO what looks to be useful information in these eleven data

Jun 16, 1995 datasets:

AUTOCICS CICS Statistics

AUTOCMDS CMD Statistics

AUTOIMS IMS Statistics

AUTOMISC Miscellaneous Statistics

AUTOMSGS MSG Statistics

AUTOOPNG OPEN Interface General Statistics

AUTOOPNS OPEN Interface Session Statistics

AUTOPORT Port Session Statistics

AUTORULE Rule Statistics

AUTOSQL SQL Statistics

AUTOVARS Subpool Statistics

Thanks to Rik van Roy, Procter & Gamble ETC, BELGIUM.

Change 13.090 A new exit, EXPDBSPU, taken upon entry to member SPUNJOBS

EXPDBSPU was added for the SAS/CPE product, which needs to rename

SPUNJOBS work datasets prior to SPUNJOBS invocation. I doubt that

Jun 16, 1995 anyone else will need to use this exit.

Change 13.089 All MXG datasets should contain variables SYSTEM, SMFTIME

TYPECIMS and for interval datasets DURATM and STARTIME, and have

VMACOMVT a dataset LABEL. These members revised for consistency:

VMACEPMV -TYPECIMS - LABEL added; dataset was labeled in VMACCIMS,

VMACORAC but I just learned that SAS does not propagate dataset

VMACTSOM LABELs when the dataset name is reused! DELETE of the

VMACMDF temporary datasets TRANSACT and CHAINED was also added.

Jun 16, 1995 -VMACOMVT - Variable DURATM added to datasets OMVTTBUF,

OMVTTVR,OMVTTCTC,OMVTTNCP,OMVTTLCL,OMVTTETE, calculated

as DURATM=ENDTIME-BEGNTIME.

-VMACEPMV - Variable SYSTEM added to all datasets.

-VMACORAC - Variables SMFTIME, SYSTEM added.

-VMACTSOM - DURATM added to TSOMCMND, DURATM STRTTIME and

SYSTEM added to TSOMDRU, SMFTIME added to TSOMDSNS.

-VMACMDF - DURATM added (note archaic MDFTRACK became the

APAF product, supported by TYPEAPAF), calculated as

DURATM=GBLENDTS-GBLBEGTS;

-VMAC78 - DURATM added to KEEP= for TYPE78PA & TYPE78SP.

Thanks to Carl E. Sommer, SAS Institute Cary, USA.
Change 13.088 Datasets TYPETALO, TYPE72SC, and TYPE78 were overlooked

BUILDPDB and were left in the //WORK file after their last use,

BUILDPD3 but are now deleted in the PROC DATASETS, like all of the

BUILD002 PDB work datasets, to minimize the work space required.

BUILD003

Jun 15, 1995

Thanks to Tracy Blackstone, Kaiser Permanente, USA.
Change 13.087 Continued enhancement to IBM-format RMF reports has added

ANALRMFR the Coupling Facility Usage Summary and Coupling Facility

Jun 15, 1995 Structure Activity; PDBOUT logic now builds all of

RMF type 7x datasets; Change 12.310 was added for a

non-partitioned machine.
Change 13.086 CICS Shutdown report failed if compressed datasets were

ANALCISH used, because POINT= cannot be used with compressed data.

Jun 15, 1995 We used POINT= only because NOBS= used to require POINT=,

but since SAS version 6, POINT is no longer required.

Unfortunately, though, without the POINT= operand, if the

SAS dataset is on tape or in tape format, NOBS=NOBS must

read the entire dataset to find NOBS. In this change, we

only needed to know if there were any observations, not

how many, so the problem was circumvented with different

logic, and compressed or uncompressed data worked fine.

7Nov02 update: Above text revised. See member VMXGOBS

to get the number of observations from

disk datasets.

Thanks to Chuck Hopf, MBNA, USA.

Change 13.085 Support for Antares' HURON SMF record creates 44 new data

FORMATS sets from the 20 subtypes of this extensive product. The

EXHRNnnn support exists now thanks to the significant contribution

IMACHURN of Glenn Bowman, who wrote the bulk of this support. I

TYPEHURN reviewed the code and made some cosmetic revisions (new

VMACHURN formats to decode bit strings, PIBx changed to &PIB.x,

Jun 15, 1995 sorted Labels), and validated with SMF data from a second

site (which uncovered an optional data segment in subtype

26), so I am quite comfortable with this code. Huron, to

be renamed ObjectStar, is a complex product with complex

instrumentation, but the "Huron Guide to Performance

Monitoring" documents much of the data quite well, and

two "Focus on Huron Performance" documents, subtitled

"Tuning Guidelines for Huron for MVS", and "Resource

Monitoring and Usage" give excellent discussions of how

best to analyze Huron data.

Thanks to Glenn Bowman, Wakefern Food Corporation, USA.

Thanks to Christa Neven, Geoep Boerenbond, BELGIUM

Change 13.084 -Support for APAR OW10393 which added a four byte field,

VMAC32 trashing the values in dataset TYPE32. The old MXG code

Jun 15, 1995 in this member did not protect for compatible changes by

IBM! The circumvention is to insert:

IF LENTSO GE 44 THEN INPUT +4 @;

after the block that inputs IOTMTOTL.

The actual change created new flag variable SMF32TCF='Y'

if variable SRMTRANS was set to zero by IBM because the

OUCB control block was not valid. When set, SRMTRANS

remains no longer valid for the duration of the session.

-Support for APAR OW12856, which adds one field (SMF32COS,

number of sections in subsequent records) to the header,

and raises the number of TSO command names that can

be monitored from 812 to 32767 by now creating more than

one type 32 record if necessary. This had no impact on

MXG logic, since MXG creates one observation for each

session segment in each record.

Thanks to Luc Mattheus, GEMEENTEKREDIET VAN BELGIE, BELGIUM.

Change 13.083 Support for TMS (CA-1) Release 5.1 (compatibly) adds

EXTMSDSN new variables to both TMS.TMS and TMS.DSNBREC. Both now