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

it clearer in the label, hence this change.

Thanks to ???, ANHYP NV, BELGIUM.

Change 12.231 Label for variable SM012CPU is now 'CPU*TIME*USED' (it

VMACARB was 'CPU TIME USED*TIMER UNITS'), rather misleading. Any

Jan 5, 1995 MXG variable with a SAS time format has already been

converted from TIMER UNITS into seconds.

Thanks to W. F. Hamilton, Scottish Widows, SCOTLAND.
Change 12.230 Weekly and monthly building of PDBs did not include new

MONTHBLD DB2 dataset DB2STATB, which is now added to both members.

WEEKBLD Note that the DB2ACCT, DB2ACCTB, and DB2ACCTP datasets

Jan 5, 1995 are NOT built in the WEEK/MONTH PDB libraries by default

(because they can be quite large, are detail, and may not

be needed - if you want then in your WEEK/MONTH PDB you

can easily add them yourself in your WEEKBLD/MONTHBLD).
Change 12.229 TELEVIEW datasets contain zero observations because the

VMACTELE initial support was written without any data to test.

Jan 4, 1995 -The INPUT of SUBTYPE should have been &PIB.1. instead of

&PIB.2.

-Replace all occurrences of '?? &NUM.2' with ' &PK.1'

so the HH,MM,SS,and TH fields are correctly input.

-Insert ?? between DATE and &PD4. in seven places.

Thanks to Tom Parquette, MONY, USA.

Change 12.228 ICEBERG status variables CAENBCUR and DFWENCUR in dataset

VMACICE ICEBRGDV are incorrect. The two statements testing each

Dec 30, 1994 of these variables for ='00'x should test for ='01'x.

Thanks to Peter McGill, NRMA, AUSTRALIA.

Change 12.227 MXG 12.03-MXG 12.05 only. VMXGHSM still fails with

VMXGHSM "BCRTBLA HAS ALREADY BEEN DEFINED AS NUMERIC". The text

Dec 29, 1994 of Change 12.112 was correct, but in member VMXGHSM you

must delete the semicolon at the end of the line reading:

LENGTH DEFAULT=4;

Thanks to Solomon Baker, The Prudential Services Company, USA.

Change 12.226 Support for Innovation Processing's IAM user SMF record.

EXTYIAM New dataset TYPEIAM provides statistics on usage and the

IMACIAM efficiency of IAM execution. This implementation keeps

TYPEIAM only the first extent location, but if users need to keep

VMACIAM all extents I will revise the MXG implementation.

Dec 16, 1994

Jan 5, 1994

Thanks to David Ehresman, University of Louisville, USA.

Change 12.225 Support for NETSPY 4.5.

EXNSPYNB -If you have LU 6.2 devices, the NSPYAPPL dataset was

FORMATS incompatibly changed because LU 6.2 devices statistics

IMACNSPY are now captured separately in these 27 new LU 6.2

VMACNSPY variables:

Dec 15, 1994 AINPUTS6 AOUTSZT6 AOUTSZW6 ARSPHOS6 ARSPNET6 ATTCHLU6

CINPUTS6 COTPUTS6 CRSPHOS6 CRSPNET6 NETRSPN6 OUTPUTN6

OVRSPPC6 SESSNO6 TOTRSPN6 TRANSNO6 T1RSPNO6 T1RSPPC6

T2RSPNO6 T2RSPPC6 T3RSPNO6 T3RSPPC6 T4RSPNO6 T4RSPPC6

USERRSP6 WRSPHOS6 WRSPNET6

and the 27 existing similarly-named (but without the "6")

variables will now contain only the statistics for your

non-LU 6.2 devices (previously they contained both).

Note that if you need statistics for totals, you will

need to decompose the ratios, sum the numerator and the

denominator separately, and then recalculate the ratio.

(VMXGSUM handles these "normalized" variables easily!).

-New dataset NSPYNCPB contains 27 variables from the new

NCP Control Block Information record, and new MXG format

MGNSPCB was created for that data.

-For all NETSPY versions, variable NSPCURVS was input as

PIB4 and thus wrong. It is now input as PIB2 and new

variable NSPCURVT='VC-S*ESTABLISHED*THIS SESSION' is

created PIB2 after NSPCURVS.

Thanks to Tim Crocker, Computer Power, USA.

Thanks to Alan Phelan, Allied Irish Bank Group, IRELAND.

Change 12.224 VM/ESA 2.2 Scheduler records cause PROBABLE DATA LOSS

VMACVMXA error message. The second occurrence of SKIP=SKIP-148;

Dec 15, 1994 should be SKIP=SKIP-100; (this is the one 67 lines after

IF MRHDRDM=2 AND MRHDRRC=5 THEN DO; /*VXSCLDDL*/

With this change, MXG tolerates VM/ESA 2.2 data records,

but there are additional fields added by IBM (1.7,1.15,

2.4,2.5,2.6,4.9,4.10) that are not yet decoded. This

text will be revised when documentation has been received

and MXG fully supports VM/ESA 2.2 records.

Thanks to Phyllis M. Parisi, University at Buffalo, USA.

Change 12.223 If the INTERVAL= parameter was entered in mixed or lower

VMXGDUR case, its value was not recognized and an error flagged,

Dec 1, 1994 but the value was then printed in uppercase, making it

quite confusing to diagnose. Now, the UPCASE() function

is used to eliminate the exposure.
Change 12.222 Datasets already using LSR are added to the report, since

ANALBLSR you should periodically review the LSR parameters to be

Dec 1, 1994 sure they are still correct. If the current buffer field

(0 when LSR is in use) is 'LSR' then LSR is already being

used. In addition, if there is a '-' beside the number

of buffers being recommended, then that signifies that a

value of 10 was forced in the report (and there were not

really 10 records in the file). This is done because the

minimum you can specify with BLSR is 10 (a lesser value

will cause a JCL error!). The '-' will usually appear on

the index component of a file. If there is a '+' next to

the number of buffers then the recommended setting

reflects the limit applied by the MEMSIZE parameter and

the EXCP count is greater than 20% of the number of

physical records in the file. In this case, sometimes

more buffers can be effective, and you may want to try

higher values. In one case, a file with 400,000 EXCPs

against the data component was reduced to 18,000 by using

BUFND=2000 and HBUFND=7500 (32m of hiperspace buffers),

and this one change reduced the run time from 2.5 hours

to 45 minutes!

Thanks to Phil Henninge, Timken Company, USA.

Change 12.221 FACOM AIM type 117 SMF record causes INPUT STATEMENT

VMACAIM7 EXCEEDED RECORD or ARRAY SUBSCRIPT OUT OF RANGE because

Nov 28, 1994 the INPUT of NUMSUBRG &PIB.4. should have been &PIB.2.

Thanks to Ian Heaney, Toyota Australia, AUSTRALIA.

Change 12.220 DB2 Package Accounting should be an SMF ID=101,SUBTYPE=1

VMACDB2 record, but that IFCID=239 record was originally written

VMAC102 as an ID=102 record. IBM tried to correct the error in

Nov 28, 1994 APAR PN56441, which did change the ID from 102 to 101,

Jan 12, 1995 but the subtype was still a zero, so APAR PN63234 now

corrects both ID and SUBTYPE. This MXG change detects

the uncorrected record and tells you you need to install

the APARs, and deletes the package record.

Pending installation of the APARs, you can use this

program to read your SMF file and to create a new

copy with the bad type IFCID=239 records corrected:

%INCLUDE SOURCLIB(VMACSMF);

DATA _NULL_;

_SMF;

FILE SMFOUT DCB=SMF;

IF (ID=101 AND SUBTYPE=0) OR ID=102 THEN DO;

INPUT @25+OFFSMF OFFPROD &PIB.4.

@29+OFFSMF LENPROD &PIB.2.

@31+OFFSMF NRPROD &PIB.2.

@;

OFFPROD=OFFPROD-3+OFFSMF;

LENLEFT=LENPROD;

%INCLUDE SOURCLIB(VMACDB2H);

IF QWHSIID=239 THEN DO;

ID=101;

SUBTYPE=1;

PUT _INFILE_ @2 ID PIB1. @19 SUBTYPE PIB1.;

END;

ELSE PUT _INFILE_;

END;

ELSE PUT _INFILE_;

Thanks to Joseph L. Schwartz, CIGNA, USA.

Thanks to Jeff Marsh, Twentieth Century Services, USA.

Change 12.219 STK ICEBERG records may cause "INFORMAT $PIB UNKNOWN" of

VMACICE "INFORMAT CHAR UNKNOWN", only if ICEBERG SMF records are

Nov 22, 1994 processed with other SMF records (e.g., if you use EXPDB

exits to add ICEBERG processing to your PDB library),

because variable VERSION in VMACICE is numeric, but it is

character in other MXG datasets and variable FLAG1 is

character in VMACICE but numeric elsewhere. Change the

variable name VERSION to ICEVERS and the variable name

FLAG1 to ICEFLAG1 and the conflict is avoided. I will

also revise my Q/A process to better detect this class of

conflicts (actually, VERSION conflict was detected but I

missed reading it - FLAG1 was not detected because it is

kept only in ICEBERG datasets and conflicted with an MXG

temporary variable, and my Q/A stream was not robust enuf

to cover that possibility!).

Thanks to Diane Eppestine, Southwestern Bell, USA.

===Changes thru 12.218 were included in MXG 12.05 dated Nov 20, 1994===
Change 12.218 This redesign of VMXGSUM now keeps only the variables

XMXGSUM that are found in the input dataset (plus those created

Nov 20, 1994 in the INCODE= logic). Previously, to prevent a failure

in the PROC MEANS, VMXGSUM created all variables that

appeared in any of its arguments, but if you used IMACDB2

to drop variables in your DB2ACCT dataset (to reduce its

size), and then you used ASUMDB2A (which invokes VMXGSUM)

to summarize, all of those dropped variables that were

listed in the SUM=, MIN=, etc., arguments in ASUMDB2A

would have been recreated (with missing values) in your

output dataset! Now you can tailor your datasets with the

_K macro in the IMAC member and use existing MXG

ASUM/TRND members to summarize only the variables that

you kept. This version only supports lists of variables

with single digit suffixes (CPU0-CPU9 works, but

CHAN01-CHAN99 will not have all 99 variables created).

The complete support for arbitrary lists will be in MXG

12.06, but none of the ASUM/TRND members in MXG contain

double-digit suffixes, so the exposure is extremely

remote. Because of this limitation, the new logic is in

a separate member named XMXGSUM, so you pioneers can help

test before I make the optimized version the default.

Simply copy XMXGSUM into your USERID.SOURCLIB library and

rename it to be VMXGSUM and you will exploit this

optimization. Note that if it ever becomes necessary to

revert to the original design, you can specify new

argument KEEPALL=YES to create all variables.

See also Change 12.233.

Thanks to Chuck Hopf, Merrill Consultants, USA.
Change 12.217 GRAFDB2 failed with TRNDDB2A data as input, because the

GRAFDB2 graphs used QWACESC (end time) instead of QWACBSC (start

Nov 20, 1994 time), which is what is used in TRNDDB2A to summarize.

All occurrences of QWACESC were thus changed to QWACBSC.

Thanks to Tom Elbert, John Alden, USA.
Change 12.216 If your requested only report PMSQL01 with PDB=SMF, the

ANALDB2R report failed (and the type of failure depended on which

Nov 20, 1994 version of SAS you were using). Requesting other trace

reports, or using READDB2 first and then ANALDB2R with

PDB=PDB will circumvent the logic error fixed here.

Thanks to Dan Null, Caterpiller Tractor, USA.

Change 12.215 Coding in progress as MXG 12.05 was to be built.

EXTY123 Support for S/390 Parallel Query Server (SPQS) type 123

EXTY123B SMF record. The segments are from DB2ACCT, but not all

IMAC123 of the fields are exactly the same, and time ran out

TYPE123 before I could finish validation. There's a remote

VMAC123 chance it will work, but I doubt it, and I do need test

Nov 20, 1994 data before I can claim support for the new accounting.

Nov 21, 1994 I listed SPQS as supported to attract test data!

Check with Tech Support before you use this support, and

this note will be revised when support is validated.

Update: Nov 21, 1994:

The MXG 12.05 support assumed there was a QWHS section,

and that header had a release value greater than 2.3, but

the IBM DSECT implies QWSH is reserved. Thus MXG may

need an inserted line after the INCLUDE of VMACDB2H with:

QWHSRELN=3.0;

to fake out the rest of the MXG code. I have revised the

VMAC123 member to remove any dependency on QWHSRELN and

streamlined the code, but structurally I still believe

that the MXG 12.05 code is okay. Again, I will revise

this text when I get my hands on a real type 123 record!

MXG 12.06 note: still no validation of this support.

Thanks to Majid Abai, Southern California Edison, USA.
Change 12.214 Analysis of Tape allocation has been revised so a tape

ASUMTALO drive is only counted once per interval. (Previously, if

TRNDTALO two jobs used the same tape drive in the same 15-second

Nov 19, 1994 interval, we counted that one instance twice, which could

cause the maximum tape drives in use to be greater than

the number of installed drives.)

You can also specify is the statistics are created for

each SYSTEM, or across all system with a new macro.

Change 12.213 New exit IMACUCB allows you to change the value of MXG

IMACUCB variable DEVICE in TYPE8911,TYPE10,TYPE1415,TYPE19,TYPE21

VMACUCB TYPE62,TYPE64,TYPE74,TYPE74,TYPE8911 members, and in data

Nov 19, 1994 set TYPE30_D observations. This is especially useful in

analyzing MXG Tape Mount and Tape Allocation statistics,

because MXG cannot tell an Autoloader from a Silo from a

Cartridge Tape device; however, you know which range of

device addresses are which, and could create values of

DEVICE='SILO' or DEVICE='AUTOLOD'

in this new exit, and then automatically your reports

from ANALTMTN,ANALTALO, etc., will be grouped and summed

by subsets of your tape devices!

Change 12.212 Support for CA/SQL's user SMF record appears to already

VMACIDMS have been in MXG - the Performance Monitor records from

Nov 18, 1994 CA/SQL are identical to the IDMS Performance Monitor's

SMF records, except that the version identifier (SMFHVER)

now contains 1200 (presumably for 12.0.0, since the prior

version IDMS 10.2.1 SMF records have the value 1021!).

Thanks to Greg Caliri, The Shareholder Services Group, USA.
Change 12.211 ACF2 variable ACPMFWHY (Reason Code) should have been

VMACACF2 FORMATed HEX2., and its label now indicates hex

Nov 18, 1994 instead of binary.

Thanks to Warren Hayward, TJX, USA.

Change 12.210 Support for Sterling Software's ASM V3.0.0 SMF type 39

VMAC39 record - their records conform to IBM's format and are

Nov 17, 1994 processed without error, including the subtype 255.

Thanks to Bob Berwick, UNIPAC, USA.

Change 12.209 Support for SMF type 116 MQM 1.1.2 Accounting record

EXTY116 creates one new dataset:

IMAC116 MQMACCT - MQM Accounting Data

TYPE116 The CPU time and counts of GETs and PUTs in four message

VMAC116 sizes (0-99,100-999,1000-9999,over 10000) are provided,

Nov 17, 1994 and the NETSNAME/UOWTIME fields are decoded from QWHCTOKN

so that specific MQM events can be matched with their

CICS or IMS transaction. Note that IMS can cause more

than one MQM record, which must be summed to accurately

provide correct totals for the IMS application.

This code has been hand checked with the hex dump printed

in the MQM System Management Guide, SC33-0806-01, which

provides a good overview of IBM's new Message Queue

Manager design and function.

Change 12.208 Support for SMF type 115 MQM 1.1.2 Statistics record

EXTY1151 creates three new datasets:

EXTY1152 MQMLOG - Subtype 1 Log Manager Statistics.

EXTY115B MQMMSGDM - Subtype 2 Message, Data Manager Statistics.

IMAC115 MQMBUFER - Subtype 2 Buffer Statistics for each Pool.

TYPE115 These statistics provide performance statistics, and the

VMAC115 Chapter 6 of SC33-0806-01 even provides guidelines for

Nov 17, 1994 interpreting and using these MQM statistics.

Change 12.207 SUBTYPE in SMF type 115 and 116 records is only one byte,

VMACSMF (instead of the normal two-byte field), so VMACSMF is now

Nov 17, 1994 modified to input SUBTYPE for this anamoly.
Change 12.206 Support for HSM APAR OW05988, which adds HSM CPU time and

VMACHSM management class name, using existing reserved fields so

Nov 17, 1994 the change was compatibly made. FSRCPUTM &PIB.4.2 now

replaces the +4 reserved area following FSRFLG2, and now

FSRMGCLN &PIB.2. FSRMGTCL $EBCDIC8. +18 replaces the +28

reserved area after FSRTRKKW. The APAR text makes no

mention of what CPU time is captured in the new field!

Variables FSRCPUTM and FSRMGTCL were added to datasets

HSMFSRST and HSMFSRTP.

Thanks to Lawrence Jermyn, Fidelity Systems Company, USA.

Change 12.205 TYPE42DS observations for VIO datasets can be recognized

VMAC42 because DEVNR=3FFFx, and most of the I/O counts/durations

Nov 16, 1994 are trash (i.e., non-zero values, but not valid either),

so now they are set missing, and new variable S42VIO is

set to "Y" to flag observations that are for VIO.

Thanks to James Purdie, Banc One Services Corp., USA

Change 12.204 For jobs which only have a type 6 SMF record, variable

BUILDPDB TYPETASK was blank in dataset PDB.JOBS. To correct, the

BUILDPD3 variable was added to the _PDB6 macro definition in

BUILD005 IMACPDB, and the variable was added to the ID= statement

IMACPDB with the PROC MEANS DATA=PRINT in members BUILDPDB,

Nov 16, 1994 BUILDPD3, and BUILD005.

Thanks to Kenneth D. Jones, SHL Systemhouse, CANADA.
Change 12.203 Support for Omegamon for CICS V300 user SMF record. The

VMACOMCI version is incompatible, as new subtypes replace old ones

Nov 14, 1994 and new variables are created in existing datasets, but

there are no new datasets. In addition, many existing

variables' formats were validated and corrected.

Now, MXG supports Omegamon for CICS Versions 550 and 551,

and Omegamon for CICS/ESA Versions 100 and 300. There

were no changes to the type 110 SMF record in V300.

Change 12.202 This CICS utility now identifies if your type 110 records

UTILCICS are actually Omegamon records (which require tailoring in

Nov 12, 1994 member IMACEXCL), which optional IBM and/or Omegamon data

segments are tacked on the base record, and which PTFs in

member IMACPTF must be updated for your sites records,

and you no longer have to go thru the detail reports and

match up fields - two simple PROC PRINTs tell all!

Note: 2010: SEE MEMBER UTILEXCL instead of UTILCICS.

Change 12.201 Support for NPM 2.2 has now been validated with real data

FORMATS for the new NPMVSaaa datasets, with these corrections:

VMAC28 -NPMVSVAD dataset: INVALID ARGUMENT TO FUNCTION INPUT and

Nov 12, 1994 variable VADSTIME is missing, because the VADSTIME=DHMS..

statement had the date and time variables reversed (1st

is VADSDTEX, then VADSTIMX). Also, variable NPMNAME was

incorrect because the two references to LSCDROFF inside

the VAD processing should have been LVCDROFF.

-NPMVSVAP dataset: VAPACTIM was trashed because it was not

documented as a datetimestamp value. VAPSECNA was trash

because it should be formatted with $HEX6. VAPDSTAT and

VAPSTATU were wrong because both had VAPCSTAT in their

percentage calculation instead of themselves.

-NPMVSVBF dataset: VBFUSER is now input as $EBCDIC8 vice

&PIB.4., and VBFAVLBF & VBFBFEXT are &PIB.2. instead of

&PIB.4. (which caused VBFPGEXT,VBFNREQ,VBFMXREQ to be

trashed, as the preceding errors messed the alignment!).

-NPMVSVDV dataset: Most variables were trashed, because I

failed to input VDVNNMAE ($EBCDIC8.) after VDVSCNAM, and

I input VDVLNKST as &PIB.4. instead of &PIB.1., and the

+4 at the end of the first input should not be there.

-NPMLUCON dataset: NCPRLSE should have been input $EBCDIC8

instead of &PIB.4, trashing NCPNEWNM and NCPGENLV.

-NPMNWCWC dataset: Most variables were trashed, because

the reserved field in the NWC section after REVL should

have been +2 instead of +1.

-All other datasets new with NPM 2.2 appear to be valid,

and the above errors had no effect on earlier NPM data.

-Minor, but some subtype values (NPMSUBTY) were not in the

MG028TY format, but now they are.

Thanks to Mr. Robbiany, Comitsiel S.p.A., ITALY.
Change 12.200 -New dataset TYPE72SC, for MVS/ESA 5.1+ in Goal Mode, is

BUILDPDB now created to keep track of the Service Classes that are

BUILDPD3 served by each MVS Server, and new variable SERVER was

EXTY72SC added to existing dataset TYPE42GO to identify the

IMAC7072 service classes that are servers. Don Deese's CPExpert

RMFINTRV is the best tool that I know that exploits all of the new

WEEKBLD data when the Workload Manager is in Goal Mode, and Don

WEEKBLDT found that these datasets enhance his ability to identify

MONTHBLD the cause of many of the delays marked as UNKNOWN in IBMs

Nov 11, 1994 new data. Give Don a call at 703 971 4497 for more info.

-The PDB-Building members were updated to create the new

PDB.TYPE72SC dataset, and in adding that dataset to the

weekly build members, I discovered that I had failed to

add several new PDB datasets to WEEKBLD/WEEKBLDT, and

also found WEEKBLD and WEEKBLDT were not synchronized.

Now, all MXG datasets built by default by JCLPDB6 will

also be built in the Weekly and Monthly datasets as well.

-MXG 12.04-only, a debugging statement IF OFFWRS GT . THEN

PUT OFFWRS= LENWRS= NRWRS=; was also deleted.

Thanks to Don Deese, (CPExpert), Computer Management Sciences, USA.

Change 12.199 Type 6 SMF record INVALID ARGUMENT TO FUNCTION INPUT is

VMAC6 caused by CA-DISPATCH SMF records that have overwritten

Nov 10, 1994 the date-part of READTIME with 'EE'x. While the real fix

is to get CA-DISPATCH corrected, MXG added an attempted

fix-up back in Change 11.342, but while the text of that

change was correct, the code I implemented from my own

change finger faulted! The two occurrences that read

SUBSTR(READCADI,1,1) must be SUBSTR(READCADI,5,1)

(because the date starts in byte 5, not byte 1!).

Thanks to Diane Eppestine, Southwestern Bell, USA.

Change 12.198 Support for InfoAccess Release 5.1 user SMF record from

EXTYODSE Network Imaging Systems Corporation adds two new datasets:

EXTYODSP ODSCMSPU - CMS (Controller Managed Storage) Report Purge

IMACODS ODSCMSRE - CMS Recache - copy report from optical to CMS

VMACODS and the ODSDEVIC optical device statistics variables were

Nov 10, 1994 completely changed and revised with durations and counts.

InfoAccess is a high-speed, high volume report storage and

retrieval system that takes character-based reports and

compresses and indexes the reports on optical platters and

in RAID arrays. Previously, the company was named Laser

Access Corporation, and the product was the LA1500 Optical

Data System, which accounts for the acronym "ODS" for this

product. The Release 5.1 records are incompatibly with