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

Inverse chronological list of all Changes:
NEXTCHANGE: Version 11
Change 11.361 The offset for MCCAVSN was hardcoded and thus wrong for

VMXGHSM some levels of HSM; now, instead of INPUT +16, the logic

Mar 25, 1994 is OFFV=65+MCCNVSNO; INPUT @OFFV ....

Thanks to Gary Matney, Twentieth Century Investors, USA.

Change 11.360 The MXG Tape Mount and Allocation Monitor is a major

ASMTAPES extension to MXG's existing MXGTMNT Tape Mount Monitor.

Mar 24, 1994 Now, both tape mounts and tape allocation-deallocation

events are recorded in SMF so you can measure how long

each tape drive was used by what job. The new monitor is

provided in ASM source code in member ASMTAPES and works

fine at two sites (one using MVS/ESA 4.3 with MIM plus

SMS, the other is at MVS/ESA 4.2), but at MVS/ESA 3.1.3

site with both MIM and SMS, the new monitor program

either waits doing nothing or ABENDS gracefully. So if

you really need this monitor now, read Change 11.101 and

then assemble member ASMTAPES (it still creates program

named MXGTMNT) and check it out. I think it is highly

likely it is ok with MVS/ESA 4.2 or 4.3, but your

feedback as to where it works and when it doesn't will

help validate for everyboth. Since both MIM and SMS get

involved in allocation, they may or may not be the

trigger, but we are actively working on the SRB dumps to

understand and fix the program for all environments.

Thanks to Bill Fairchild, Royal Associates, USA

Thanks to Chuck Hopf, Primerica, USA
Change 11.359 If you modified the interval in ASUMTMNT, GRAFTMNT will

GRAFTMNT not correctly place the points on the graphs since it

Mar 24, 1994 was using the HOUR of the time as the axis. It now uses

the time at 3600 second intervals.

Thanks to Chuck Hopf, Primerica, USA
Change 11.358 ASUMHSM, TRNDHSM, GRAFHSM provide some ability to report

ASUMHSM on dataset movement caused by HSM. Since this can be a

TRNDHSM significant contributor to batch run times as well as TSO

GRAFHSM response, you may find these summarization, trending, and

Mar 24, 1994 graphical analysis of HSM useful.

Thanks to Chuck Hopf, Primerica, USA

Change 11.357 The second pair of variables named LLSNAFB/LLSNAFN are

VMACNSPY now named LLSNAEB/LLSNAEN, and LLSNAEB was added tothe

Mar 24, 1994 MGBYTES format (the "E" vars are for SNA over Ethernet).

Thanks to Warren Hayward, TJX, USA.

Change 11.356 Change 11.352 was revised after new iterations. Type 35

VMACOPC caused ABEND that was fixed, and subtype 'G' is now

Mar 24, 1994 supported. There are some spanned subtype 24 records

that MXG does not yet handle correctly; at present all I

could do was to recognize I missed a spanned record with

a message on the log; this only affects the OPC24xxx

data sets, and will be fixed soon. Fax if you need it.

Thanks to Randy Shumate, Mead Data Central, Inc.

Thanks to Maureen Walshe, IBM Nordiska Laboratoirer, SWEDEN.
Change 11.355 Change 11.351 was revised after the Mar 23 early tapes

BUILDPDB were sent. The WEEKBLD/WEEKBLDT/MONTHBLD members had not

Mar 24, 1994 been revised until Mar 24.
Change 11.354 CICS Statistics variables A21LUTTM and A21SNTTM were not

VMAC110 correct; a real value of 30 minutes was reported as only

Mar 24, 1994 .03 seconds.
==Changes thru 11.353 were in MXG 11.11 created March 23, 1994===
Change 11.353 LEGENT SAR records had a number of fields added in 1993

VMACSAR that are now supported in MXG. The maintenance has no

Mar 23, 1994 version/release, only Change #05- Change #9 in DSECT!

Thanks to Bob Mattingly, ARCO-EIS, USA.

Change 11.352 OPC records caused INPUT STATEMENT EXCEEDED RECORD LENGTH

EXOPC24X for TRLRCTYP=35 (delete all references to TRLPOS35). For

IMACOPC TRLRCTYP=24 MT0TYPE=9 (expand IF MT0TYPE NE 6 THEN to

VMACOPC IF MT0TYPE NE 6 AND MT0TYPE NE 9, and expand 7 LE MT0TYPE

Mar 23, 1994 LE 8 to 7 LE MT0TYPE LE 9). Support was added for the

Mar 24, 1994 MT0TYPE='G' (MT0TYPE is a number, so character 'G' =199)

which creates new dataset OPC24_G (not to be confused

with existing dataset OPC24D_5, which caused the new exit

and dataset macros to be EXOPC24X for OPC24_G dataset).

OPC support was revised after March 23 tapes shipped.

Thanks to Randy Shumate, Mead Data Central, USA
Change 11.351 RMF dataset TYPE77 is now automatically created by MXG's

BUILDPDB BUILDPDB/BUILDPD3 algorithms. TYPE77 reports ENQUE

BUILDPD3 conflicts and delays, and is expected in ANALRMFR for

BUILD001 replication of IBM RMF reports from MXG datasets. Logic

BUILD003 in WEEKBLD/WEEKBLDT/MONTHBLD now also expectes the TYPE77

BUILD518 dataset. Added Apr 18: This change is INCOMPATIBLE if

BUILD606 you have tailored BUILDPDB to add TYPE77 processing. See

WEEKBLD text of Change 12.013. This change was also incomplete

WEEKBLDT for JES3.

MONTHBLD

Mar 24, 1994
Change 11.350 Additional CICS Shutdown Reports were added. One problem

ANALCISH with Mode Table report (many pages, all zeros) with all

Mar 23, 1994 APPLIDs went away when a single APPLID was reported, but

this will be investigated as soon as data tape received.

Also, Last Reset time is different between IBM & MXG.

Thanks to Neil Ervin, Huntington Bank Service Company, USA.

Change 11.349 Variable SHEETPRN is now automatically added to PDB.PRINT

IMACPDB to count sheets printed.

Mar 22, 1994

Thanks to Jill Hansen, South Dakota Education, USA.

Change 11.348 Variable DVLNUCBA in dataset DCOLVL is now formatted as

VMACDCOL HEX8 and is LENGTH 5 (because it is numeric, five bytes

Mar 22, 1994 are required to store all possible hex digits).

Thanks to Al Rozewski, Parker Hannifin, USA.

==Changes thru 11.347 were printed in MXG Newsletter 25 dated 26Mar94===
Change 11.347 Support for Boole & Babbage CICS/Manager Statistics data

EXCICBBD in type 110 SMF record, subtype BB02 (which MXG sets back

EXCICBBF to SUBTYPE=2 for processing), Statistic STIDs:

EXCICBBG STID DATASET DESCRIPTION EXIT MEMBER VARS

EXCICBBL 200 CICSBBSI SIT EXCICBBS 52

EXCICBBR 201 CICSBBRC RCT EXCICBBR 47

EXCICBBS 202 CICSBBLT LT X EXCICBBL 55

FORMATS 203 CICSBBFC FCT EXCICBBF 57

IMACCICS 204 CICSBBGL GLOBAL PERFORMANCE EXCICBBG 37

IMACICBB 205 CICSBBDL DLI EXCICBBD 44

VMAC110 Those 6 datasets are created, but they will have zero obs

Mar 20, 1994 and only 15 variables unless you enable processing - see

member IMACICBB for enablement procedure and comments.

Adding this support uncovered several errors in field

alignment (three sets of CPU fields) that will be fixed

by Boole's PTF BPC2312 which you must request and install

for that dataset to be valid.

Thanks to ???, VW Wolfsburg, GERMANY.

Change 11.346 Netmaster 2.2 added new variable SMFNCUSR to type 39 data

VMAC39 that is now decoded and added the TYPE39 datasets. This

Mar 19, 1994 is the only reported change in Netmaster 2.2 records.

Thanks to Colin Bowen, Old Mutual, SOUTH AFRICA.

Change 11.345 TCP/IP addresses contained blanks when only one digit was

VMACTCP used for a node; now the blanks are stripped out by using

Mar 19, 1994 TELLOCAL=COMPRESS(TELLOCAL); on all addresses.

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

Change 11.344 Support for CADAM V3R2 Statistical Data plus corrections

VMACCADM to MXG were provided by this user enhancement. See the

Mar 18, 1994 excellent notes at the beginning of the member.

Thanks to Jouke van Schepen, Fokker Aircraft BV, NETHERLANDS.

Change 11.343 NPM Type 28 NPMLANOD dataset (added in NPM Version 2) did

VMAC28 not decode CSL section correctly, causing INPUT STATEMENT

Mar 18, 1994 EXCEEDED with NPMSUBTY='A0'x, which is 4-bytes shorter

than the 'A1'x subtype. The four final fields LCSLRPTO-

LCSLRSFR are now input only for NPMSUBTY=0A1X and three

new variables LCSLPDUD,LCSLMFRD,LCSLURFR are instead

input for NPMSUBTY=0A0X.

Thanks to Pat McGuire, Texas Instruments, USA.

Change 11.342 CA-DISPATCH 5.1 corrupts READTIME in TYPE6 records - the

VMAC6 date can be 1-2 days in the future! CA stores a 01x in

Mar 18, 1994 the 1st byte of READTIME as a flag. A real read time of

Sep 22, 1994 00021CE8x (00:23:04) is corrupted to 01021CE8x (47:59:16)

and those 48 hours are added to midnight of read in date!

My guess was that CA decided that since 0083D600x is 24

hours, they could use the 1st byte of time for DISPATCH,

(just like CA uses the first byte of date for CA7), but

that is not the case; CA now acknowledges that READTIME

field is being corrupted and CA Level 2 is working at the

one reporting site to develop a fix. I had already added

protection in MXG 11.11 to reset the first byte to zero:

Replace READTIME SMFSTAMP8. @;

with READCADI $CHAR8. @;

IF SUBSTR(READCADI,5,1) GT '01'X THEN

SUBSTR(READCADI,5,1)='00'X;

READTIME=INPUT(READCADI,SMFSTAMP8.);

but now the site reports the READTIME is off by 2 hours,

so it appears the time part of READTIME is just bad,

until CA develops a fix.

See Revision by Change 12.199; 1,1 changed to 5,1.

Thanks to Giovanni Dossena, Einchem Elastomeri S.R.L., ITALY.

Change 11.341 TYPE94 variables SMF94Axx were labeled as EJECT when they

VMAC94 should are AUDIT, and variable SMF94EIN/SMF94EPM were

Mar 17, 1994 dropped from the KEEP= list, as they do not exist, and

could be confusing!

Change 11.340 Support for STC ICEBERG 9200 Disk Array Storage Subsystem

EXICECHA creates four datasets, one for each subtype of the SMF

EXICEDEV interval record which are provided by StorageTek:

EXICEDRV ICEBRGSY - Capacity and Space Utilization - per Subsys

EXICESYS ICEBRGCH - Channel Interface Statistics - per channel

IMACICE ICEBRGDV - Device and Its Cache Statistics - per device

TYPEICE ICEBRGDR - Drive Module Statistics - per drive module

VMACICE The range and content of the ICEBERG statistics are quite

Mar 9, 1994 impressive & comprehensive for this new technology, with

utilization counts and durations provided. This support

has been syntax checked, and simulated test data has been

processed, but no real-world users have used the data yet

Change 11.339 MXG 11.08 thru MXG 11.10. the last 380 lines of this RMF

ANALRMFR Report member were inadvertently deleted. The lines were

Mar 7, 1994 restored in MXG Early 11.11, without a Change number.

Thanks to Norbert Korsche, OMV-AG, AUSTRIA.

==Changes thru 11.338 were in the MXG Version Early 11.11 dtd Mar 8, 94=
Change 11.338 APAR UW04108 adds new variables to TYPE42 datasets:

VMAC42 TYPE42SR: ICLS RCLS SEQIOS

Mar 7, 1994 TYPE42DS: ICLS RCLS SEQIOS

Sequential I/Os are now counted separately (SEQIOS) and

are NO LONGER included in cache statistics (candidate

I/Os and hits). RLCS are Record Level Caches, ICLS are

Inhibit Cache Loads. ICLS only include those ICLS that

are set by DCME, not those by the STARTIO driver.

An additional APAR due out later this year adds even more

valuable instrumentation to the data set statistics:

TYPE42DS: BLKSIZE DEVNR STORCLAS VOLSER

S42AMDRB S42AMDRR S42AMDWB S42AMDWR

S42AMSRB S42AMSRR S42AMSWB S42AMSWR

S42AMZRB S42AMZRR S42AMZWB S42AMZWR

The long-needed VOLSER and DEVNR will be the first volume

for most multi-volume or striped datasets; however, for

sequential SAM access, there will be one record per

volume. The new Access Method fields (S42AMxxx) are

generated only for access methods that support DCME; the

new S42AMxxB variables count blocks read/written for

sequential/direct/directory and the S42AMxxR fields will

contain the corresponding I/O delay time (they are not

yet implemented). The directory counts do not include

STOW or BLDL yet, and there's more design ongoing to

capture as much as possible (eg., VIO and PDSEs). Note

how IBM is making life simple for us both, providing me

with early documentation so this support is already in

place in MXG 11.11 for when you get the APARs!

Thanks to Jeff Berger, IBM SSD, USA.
Change 11.337 CA's TMS can cause a type 80 (RACF) record to be created

VMAC80A for BLP processing, (a ZAP from CA is required to enable

Mar 7, 1994 creation of the records), but they exposed an MXG design

error: variable RESNAME was blank and variable OLDDSN

contained resource name. MXG now correctly inputs the

Resource Name into variable RESNAME; I should then store

RESNAME into OLDDSN only if RACFEVNT=04 (a RENAME), but

since OLDDSN always has contained the Resource Name, and

since you should not have to change your reports, I chose

to continue to put Resource Name in both RESNAME and

OLDDSN variables.

Thanks to Simon Hendy, Reader's Digest European Systems.

Change 11.336 Boole & Babbage CMF PTF BPM4681 adds new variables to

VMACCMF these existing datasets:

Mar 6, 1994 CMF27C93 C279WEH,C279WER,C279WFM

CMF27CSD CMF27CHN,CMF27CU2,CMF27DEV,CMF27LCU,

CMF27MDR,CMF27OBR,CMF27UA1,CMF27UA2,

CMF27uty, and CMF27VOL (the VOLSER!)

In the CMF27CSD dataset, the existence of the new fields

can be identified by testing CMF27VOL; if it is non-blank,

the record was created after PTF BPM4681.

Thanks to Matthew McCue, United Parcel Service, USA.

Thanks to John Piccone, United Parcel Service, USA.
Change 11.335 A minor correction to the revised VMXGSUM summarization;

VMXGSUM if the first data step was not required by the SORT, the

Mar 6, 1994 PROC MEANS looked for MXGSUM1 when it wanted MXGSUM2;

also, a specious error message when the length of the

INDATA= string was less than 40 bytes was eliminated. It

needs to be stressed that the changes made to VMXGSUM are

INCOMPATIBLE if you have tailored any of these members:

ASUMDBDS ASUMDB2A ASUMDOS ASUMHPCS ASUM70PR

DAILYDSN GRAFDB2 GRAFLPAR TRNDDB2A

You must retrofit your tailoring, starting with the new

member in MXG 11.11 (see the text of Change 11.309).
Change 11.334 Batch LSR for VSAM can produce incredible savings, by

ANALDSET using memory for buffers instead of repetitive I/O to the

ADOCBLSR same record. Jobs cost less, use less CPU, fewer I/Os,

ANALBLSR and run in tens of minutes instead of tens of hours. This

Mar 6, 1994 new analysis by Chuck Hopf adds new variables in existing

ANALDSET program (that reads SMF and combines type 14/15,

type 64, and type 30 data) to its output dataset DSETOPEN

which is then used as input to ANALBLSR's algorithms to

identify the jobs and VSAM files that could benefit from

BLSR. ANALBLSR also reports any existing Batch LSR usage

and will suggest increase or decrease in buffering where

appropriate. Implementing Batch LSR requires no change

to the application; only a simple JCL change is required,

and example JCL is in member ANALBLSR. Chuck's full

research paper on this timely subject will be in member

ADOCBLSR when it is available. Chuck points out that for

random access to the same records/index, increasing the

number of buffers (BUFNI,BUFND) does not eliminate I/O.

You would expect that if the data was in the buffer VSAM

would find it there, but actually without Batch LSR, I/O

is done instead of lookaside into the buffers! One case

of an Index with only 6 records had one million EXCPs for

a single step; using BLSR with 10 buffers reduced the I/O

count to seven! ANALBLSR lets you set thresholds of the

amount of memory you want to use, and the percentage of

the total I/O for the step, before it will be selected as

a candidate for Batch LSR, and is self-documenting. This

is still ongoing research.

Thanks to Chuck Hopf, Primerica, USA.
Change 11.333 ANALDB2R PMAUD02 Authorization Failure report had N/A for

ANALDB2R table/object name, when there should have been a name.

DIFFDB2 The length of a SUBSTR() was incorrect, causing tests for

IMACDB2 character values to be incorrect. The DB2PM manual was

READDB2 used to decide when a Target/Owner is printed, and it

Mar 6, 1994 says that they are not printed for "ARCHIVE", yet their

report does print it, so we revised our logic to match!

ANALDB2R PMSTA01 Statistics report timestamps printed

were unclear or misleading. There are two sets of

timestamps; the first is the time range of the data that

was read, the second is the range of data summarized on

that page, if INTERVAL= is specified. Also, DIFFDB2,

IMACDB2, and READDB2 were corrected to use _LDB2STA

instead of the hardcoded PDB.DB2STATS, and IMACDB2 and

READDB2 now know about the new DB2 3.1 dataset DB2ACCTP.

Thanks to Wai Choong Mak, Development Bank of Singapore, SINGAPORE.

Change 11.332 Support for Candle's Omegamon II for SMS user SMF record

EXOMSMDV creates two new datasets:

EXOMSMJB OMSMSDEV - DASD Device Statistics

IMACOMSM OMSMSJOB - JOB and DSNAME activity on each volume.

TYPEOMSM This code has been tested with actual data, but has not

VMACOMSM been extensively validated by real users, yet!

Mar 5, 1994
Change 11.331 The contributed RACF reports program WPDBRACF had to be

ANALRACF changed due to an apparent change in the way that some

Mar 4, 1994 formatted values were named in the PROC TRANSPOSE. The

RENAME= list for dataset RACFREP2 was revised.

Thanks to Neil Campbell, Inland Revenue, ENGLAND.
Change 11.330 DB2 Audit Detail report, Completion Code, was incorrect,

ANALDB2R causing "INVALID NUMERIC DATA" message on the SAS log.

Mar 4, 1994 Two tests for QW0083AD=0 and two tests for QW0087AD=0

should have tested for hexadecimal character zero instead

of numeric. The two pairs of statements now reading :

IF QW0083AD=0 THEN .... and IF QW0087AD=0 THEN ....

must be changed to read:

IF QW0083AD='00'X THEN ... and IF QW0087AD='00'X THEN ...

Thanks to Wai Choong Mak, Development Bank of Singapore, SINGAPORE.
Change 11.329 CICS/ESA DFHSTUP Shutdown Statistics Reports can now be

ANALCISH printed by MXG, either from a raw SMF file, or from a PDB

VMAC110 library (with minor modifications to BUILDPDB). This is

Mar 4, 1994 a significant contribution that uses ESA CICS datasets to

replicate the important IBM Shutdown reports. You can

// EXEC MXGSAS

//SMF DD DSN=YOUR.SMF.TYPE110.records,disp=shr

%ANALCISH(PDB=SMF);

to generate all reports from raw SMF data. You can also

generate these reports regularly, from your PDB, but you

will need to tailor BUILDPDB so that it copies all of the

CICS statistics datasets from the WORK file into the PDB.

You must add, in member EXPDBOUT, this code:

PROC COPY IN=WORK OUT=PDB;

SELECT CIC:;

and then you can invoke %ANALCISH(PDB=PDB); to print

shutdown reports for all CICS regions. Additional macro

arguments let you select date/time/region, and to select

only the desired report.
An minor error in VMAC110 was also corrected; member

IMACCICS is now included by its VMAC, instead of in its

TYPE member or by BUILDPDB. This clerical oversight only

affected me when exploiting my new "_L" logic, but should

have no effect in the field!

Note for the experts: I needed to do this so that I

could null out the CICSTRAN data set (which has high

volume, and is not currently used by ANALCISH) when

I ran against SMF data, and the mislocated %INCLUDE

did not let me. Normally you would null out a dataset

by EDITing the product's IMACxxxx member and change

its "_L" macro's dataset name to "_NULL", but you can

also null out any MXG dataset on the fly, without EDIT

of the IMACxxxx member, by using this syntax:

%INCLUDE SOURCLIB(VMACSMF,VMAC110);

MACRO _LCICTRN _NULL_ %

DATA _VAR110; _SMF; _CDE110;

(You must be at SAS 6.08 for the _NULL_ operand to be

a valid argument of the OUTPUT statement!)

In addition to producing the CICS Shutdown Report, member

ANALCISH lets you see what variable from what MXG dataset

is used for what report field, by reading the code! This

set of reports has been long overdue; the most important

reports have been implemented for both CICS 3.2 and 3.3,

but there are more reports (especially the detail reports

by transaction) that were not finished in time for 11.11.

Thanks to Willi Weinberger, Gothaer Versicherungsbank, GERMANY.

Thanks to Siegfried Trantes, Gothaer Versicherungsbank VVAG, GERMANY.

Change 11.328 Division by zero if there were no TSO transactions in an

TRNDRMFI interval. Change PCTTRIV=TRIVTRAN/TSOTRAN*100; to read

Mar 4, 1994 IF TSOTRAN GT 0 THEN PCTTRIV=TRIVTRAN/TSOTRAN*100;

Thanks to Norbert Korsche, OMV-AG, AUSTRIA.

Change 11.327 MXG 11.09A and 11.10 only. Change 11.270 caused INVALID

TYPEMON8 DATA FOR TIAPREQ in the MONISYST dataset if the Landmark

Mar 4, 1994 record was from 8.1 with an archaic history segment (i.e.

when LENGTH=2276). The test in MXG for IF LENGTH GE 1464

should have been IF LENGTH EQ 1464. (The error did not

affect the MONITASK dataset, and the history segment is

no longer created by Landmark.)

Thanks to John Goodstat, Gardner Merchant, ENGLAND.

Change 11.326 Sterling's NDM has been renamed to Connect Direct 1.7.01

VMACNDM and the format of the PT record changed, causing INVALID

Mar 4, 1994 DATA for HH messages. Replacing the single line reading

Mar 21, 1994 INPUT +30 with LOC=LENGTH-21;INPUT @LOC corrected some

records, but there are "PT" records with invalid values

for date/times of 000000000001000Ax & 2800000000FC5B10x

that I need to talk to Sterling about, but I can't find

anyone there to return my call, and I only have the PT,

CT, & MC segments corrected thus far, and I still have

no response from Sterling. If you need to process NDM

records now Connect Direct, send us a fax request, ande

we will advise you of the current status.

Thanks to John Goodstat, Gardner Merchant, ENGLAND.
Change 11.325 Type 42 subtype 6 read from VSAM SMF caused STOPOVER.

VMAC42 (There was no error when dumped BSAM SMF was read.)

Mar 1, 1994 Calculation of these three offsets did not include the

"+OFFSMF" at the end of the line. They should read:

OFFJDDSO=OFFJDDSO-3+OFFSMF;

OFFDSIOO=OFFDSIOO-3+OFFSMF;

OFFJDDSO=OFFDSNXT-3+OFFSMF;

The GMT conversion algorithm should also be changed to:

GMTOFF42=100*FLOOR((SMFTIME-SMF42PTE+10)/100);

Thanks to H. Placht, RWD Gmbh Datenverarbeitungsgesellschaft, GERMANY

Change 11.324 Variables SAMPSKPD, RMFIIIRC and INTRVSYN were always

VMAC7072 blank, because variable CONVFLAG should have been input