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

only 601MB (megabytes)! So much for IBM consistency in

use of "MB". Because VMACDCOL assumed MB was megabytes,

it multiplied by 1024 to convert to bytes for MGBYTES, so

this change replaced all occurrences of 1024 with 1000.

The unknown UK user also noted that the test that set the

values of DCVEVLCP, DCVEBYTK, and DCVELSPC should have

tested DCVERROR instead of DCVFLAG1.

Thanks to Derek Cespedes, Florida Power and Light, USA.

Thanks to ???, ???, ENGLAND.
Change 08.284 Major revision of support for EPILOG/CICS was finished

EXEPCIJC after Newsletter NINETEEN went to the printer. The change

EXEPCISF is INCOMPATIBLE with previous MXG EPILOG/CICS support, as

EXEPCISI the TYPEEPIL dataset no longer is created. Instead, these

five new MXG datasets are created by TYPEEPIL execution:

EXEPCISU EPILCITR - Transaction data - most important.

EXEPCITR EPILCISU - Startup.

TESTOTHR EPILCISF - Suffix.

VMACEPIL EPILCIJC - JCL Parameters

Apr 4, 1991 EPILCISI - SIT Parameters

Each dataset keeps only the appropriate variables.

VMACEPIL code was restructured, with a separate section

for Releases 450-451, which is different from 440 format.

Only EPILCITR records were available for testing from 450

but there is no change in the format when you install 451

so I anticipate no problems. While the (archaic) 440 code

might need work, Candle solves 440 problems by sending a

copy of 451!

Thanks to Richard Evans, Mervyn's, USA.
==Changes thru 8.283 were printed in MXG Newsletter NINETEEN==========
Change 08.283 SAS 6.06 errors like "Data set is not sorted" that were

SAS 6.06 discussed under Usage Note 1000 were finally resolved on

Mar 29, 1991 the March SAS Usage Notes Tape by the inclusion of SAS's

ZAP Z6062141. If you do not have the March SAS 6.06 tape

installed (it's now a 100% pre-applied ZAPed loadlib in

IEBCOPY format), you MUST at least download and install

Z6062141.
Change 08.282 This new member documents all of the "Products" that MXG

PRODUCTS supports. PRODUCTS identifies which product versions are

Mar 29, 1991 supported by which version of MXG and the member name of

the MXG member(s) that processes that product's records.

Use your online editor to search for the product name to

find the entry for a specific product, which will contain

the names of MXG datasets created, their exit names, etc.

Suggestions for enhancement are most welcome.

Change 08.281 This new member documents the IMAC.... exit members which

IMACAAAA tailor MXG for your site. Some IMACs (IMACSHFT,IMACWORK)

Mar 29, 1991 should always be tailored, but most IMACs are not needed

to be changed by most sites. This new member groups the

different purpose IMACs together to make it easier for an

installer to know which IMACs affect their site. The IMAC

itself contains self-documenting comments for its use.
Change 08.280 Transit time reports from DB2 trace are now in ANALDB2R.

ANALDB2R Selection criteria is more robust and is uniform across

Mar 29, 1991 all reports, and can include SHIFT or DB2 Trace Number.

ANALDB2R reports can be produced from either the PDB or

the new TRNDDB2x data (See Changes 8.276 and 8.279) for

Accounting and Statistics reports. Data from the TREND

databases can be input along with data from your daily

PDB to compare history with current with PDB=PDB TREND

syntax to cause ANALDB2R to combine both data sources.

Transit time and SQL trace activity analysis were added

by ANALDBTR (Change 2.249) which creates the trace-pair

data sets needed. ANALDB2R now will use those data for

these new reports, and for the I/O Activity Summary, and

eventually all DB2 trace reports will be restructured to

use the ANALDBTR trace-pairs.
Change 08.279 Summarization of DB2STAT0 and DB2STAT1 statistics data

TRNDDB2S into the TREND database uses PDB.DB2STAT0 and DB2STAT1

Mar 29, 1991 as input to create TREND.TRNDDB20 and TREND.TRNDDB21.

Summarization is by WEEK by SHIFT within SYSTEM, and DB2

Sub-System ID, QWHSSSID. Note that Version 8.8 ANALDB2R

can use these new TREND datasets as input for reports.

See DOCTREND for the structure of MXG trending.
Change 08.278 This significant user contribution for AS400 Records will

AS400PDS eventually be a fully supported part of MXG, but time ran

Mar 28, 1991 out for development and testing. This member is actually

a 27-member unloaded PDS, and is provided so that leading

edge sites who would otherwise have had to start from the

beginning, may find this a starting point for AS400 data.

It should be regarded as preliminary, unverified, and is

certain to change in structure. Your input is welcome.

Thanks to Carolyn Barnett, Sentara Health System, USA.
Change 08.277 Variables INAVG and READYAVG were added to the RMFINTRV

RMFINTRV dataset. I had actually planned to do further revisions

Mar 28, 1991 to RMFINTRV for MVS/ESA, but ran out of time!

Thanks to ???, ???, EUROPE.

Change 08.276 Summarization of DB2ACCT dataset into the Trend Database.

TRNDDB2A Input is weekly PDB.DB2ACCT into TREND.TRNDDB2A. The time

Mar 28, 1991 summarization is for each shift for each week within the

SYSTEM SHIFT QWHSSSID QWHCPLAN QWHCCN values.

Both ANALDB2R and/or GRAFDB2 will produce Accounting

Detail and Summary reports and graphs from TRNDDB2A.

See documentation in the member itself.
Change 08.275 Graphical analysis of DB2 data from MXG PDB or TREND

GRAFDB2 data libraries (three sets of graphs) can be produced

Mar 28, 1991 and a graphic catalog (default DDname of PDB) is built.

See comments in member for documentation.

Change 08.274 Documentation of the techniques implemented in MXG use

DOCGRAF of SAS/GRAPH, and briefly describes the MXG GRAF....

Mar 28, 1991 members and the graphic catalog they produce.
Change 08.273 New style macro generates the code needed to download

VMXGDOWN all of the datasets contained in a single SAS data

Mar 28, 1991 library. This utility is the logical equivalent of:

PROC DOWNLOAD DATA=PDB._ALL_ OUT=PCPDB;

which does not exist.

Thanks to Chuck Hopf, Hopf Consulting, USA.

Change 08.272 The MXG utility UTILPRAL will print all datasets in a

UTILPRAL SAS data library. This change simplifies that process

VMXGPRAL by executing as a single step with no external DD.

Mar 28, 1991

Thanks to Chuck Hopf, Hopf Consulting, USA.
Change 08.271 Graphics table added samples for IBM3825 and IBM3827

VMXGGOPT devices (but only in portrait mode; rotating requires the

Mar 28, 1991 use of templates, and is non-trivial). See SAS Technical

A SAS note xxxx that discusses rotation. Also, standard

PATTERN/SYMBOL statements were added for consistency so

they can (eventually) be used in all GRAF.... members.

Change 08.270 Printer analysis logic for NPRO was corrected and support

ANALPRTR for XEROX XPAF CPU calculation (based on testing of early

ASUMPRTR releases of XPAF) was added. The default value for NPRO

IMACPRTR is now zero, since only the (dead?) 3800-3 used NPRO.

Mar 28, 1991 For sheet printers, PAGECNT is equal to the number of

physical sheets that went through the printer; SHEETPRN

is the number of sides printed. For the printer THRUPUT

calculations, SHEETPRN is now used instead of PAGECNT if

SHEETPRN is non-zero (using PAGECNT, duplexing made the

printer's thruput half-speed)!

Thanks to Chuck Hopf, Hopf Consulting, USA.
Change 08.269 Preliminary support for log produced by IBM's TPNS. Its

TYPETPNS pretty simple, but there is not a whole lot of data.

Mar 27, 1991
Change 08.268 Because it was not always clear on the log when %VMXGSUMs

VMXGSUM invocation actually began, a RUN; statement was inserted

Mar 27, 1991 after line 019500.
Change 08.267 Protection for invalid accounting field lengths has been

IMACACCT enhanced. The total length of accounting data was always

Mar 27, 1991 compared with record length, but individual account field

lengths were not verified, until now. A site with a badly

corrupted type 30 record (due to an error in their SMF

exit code) ABEMDED with a STOPOVER. With this change, the

unnecessary STOPOVER is avoided, the bad record detected,

and an MXG message is now printed on the log. The site

chooses to remain unknown!
Change 08.266 Thanks for IBM's excellent vendor support in the CICS/ESA

Mar 27, 1991 Early Test Program, we can confirm that not only will

MXG Version 8.8 correctly process CICS/ESA 3.1 SMF data,

but also that it will not fail with CICS/ESA 3.2 records!

Change 08.265 Support for NETSPY Release 4.0 was added by this change.

EXNSPYAC Four new variables were added to the NSPYAPPL dataset,

VMACNSPY eight new variables were added to the NSPYVIRT dataset,

Mar 26, 1991 and the new NSPYACCT data set with 68 variables for the

Network Accounting type 'C' NETSPY record is created.

Thanks to Richard Warren, Solomon, Inc, USA.

Change 08.264 This change summarizes and documents the MXG DASD space

ANALVVDS management facilities that have been enhanced in 8.8.

ASMVVDS Member JCLDASD/JCLDASD6 contains the sample JCL that is

FMXGUCBL suggested to allow you to process your VTOCs and VVDS.

IMACVVDS Comments in the MXG members elaborate on each of the

JCLDASD programs discussed. Note that "ASM" members are written

TYPEVVDS in IBM Assembler and must be assembled and link edited

VMXGVTOC before they can be used. Sites which have installed

VMXGVTOF DFP 3.2 or later should first look at IBM's DCOLLECT

VMXGVTOR utility (supported by MXG's TYPEDCOL member) to measure

Mar 26, 1991 their DASD farm, as it may be adequate for most sites.

The VTOC/VVDS processing described below gives more data

that is presently provided by IBM's DCOLLECT.
1. VTOC Processing

Original VTOC processing in MXG consisted of these three

members, which in concert would dynamically allocate all

DASD volumes that were online, read each VTOC, and build

three data sets describing VTOCs, Datasets, and extents

in your DASD farm:

FMXGUCBL - a function for dynamic allocation, used by

VMXGVTOR - a %macro that interated the execution of

VMXGVTOC - a SAS program that read a single VTOC

Execution of this sequence created the three MXG datasets

to be used for DASD space management, chargeback, etc:

VTOCINFO - One obs per VTOC, describes the volume

VTOCLIST - One obs per dataset, describes each dataset

VTOCMAP - One obs per extent, for pack analysis.

Large sites (over 500 volumes) found that VTOC reading

with SAS itself, while functional, took scores of seconds

per VTOC. Philip Morris wrote their own ASM program and

graciously contributed it to MXG, and cut the VTOC read

time to seconds per VTOC! You need only to assemble and

link edit the ASM program:

ASMVTOC - "Fast" VTOC reading assembly program

and then execute PGM=ASMVTOC to read all VTOCs and write

a single flat file to the VTOCDUMP DDname, which is then

read by the MXG SAS %macro program %VMXGVTOF in member

VMXGVTOF - Reads VTOCDUMP file created by ASMVTOC and

create three VTOC.... data sets.

(The example JCL in JCLDASD/JCLDASD6 then copies the

three VTOC.... datasets to the SAS data library that

is pointed to by the MXGDASD DDname.)

This is the recommended method for reading all VTOCs.

2. VVDS Processing.

Originally VVDS described only VSAM dataspace on a volume

but with SMS, the VVDS contains an entry for both VSAM

(in a VVR in the VVDS), and for non-VSAM (in a NVR in the

VVDS). Keeping track of VSAM data spaces on a volume was

the primary reason that the MXG Assembly program

ASMVVDS - read all VVDS and create a user SMF record

which was then processed by MXG members

IMACVVDS - identifies the SMF record number you chose

TYPEVVDS - to read the SMF VVDS records and create

the MXG dataset of of the same name with

an obs for each entry in the VVDS.

Additional idiosyncrasies in VVDS entries were found and

corrected in analysis which is now inclued in member

ANALVVDS - executes IMACVVDS/TYPEVVDS, and adds logic

to "clean-up" the VVDSf data.

Member ANALVVDS creates these two MXG datasets for VVDS:

TYPEVVDS - One obs per VVR, cleaned up

VSAM - One obs per VSAM data set.

The two datasets are written to the DDname of MXGVVDS.

Change 08.263 Three variables added by DB2 2.2 were not deaccumulated.

DIFFDB2 Variable Q3STRDON in DB2STAT0 and variables QTAUCCH and

Mar 26, 1991 QXMIAP in DB2STAT1 are now DIF()'d in DIFFDB2.

Thanks to Siegfried Trantes, Gothaer Versicherungsbank VVAG, GERMANY.

Change 08.262 Support for TSO/MON Release 5.3.0 has been added and has

EXTSODRU been tested, due to excellent support from LEGENT; not

VMACTSOM only did this vendor provide documentation in advance of

Mar 25, 1991 the general availability of the product, but also sample

SMF records were provided for testing! The major change

is the new TSOMDRU Dynamic Resource Utilization dataset,

which allows you to identify the most resource intensive

commands so that you can then enable detail recording for

those commands. Minor changes included new fields in the

TSOMSYST and TSOMCALL datasets containing Hiperspace page

in and page out counts.
Change 08.261 TSO/MON system records are restricted to 4089 bytes. When

VMACTSOM more users are logged on than fit in 4089 bytes, TSO/MON

Mar 25, 1991 writes multiple records, which MXG recognized, but the

INTRVTM value was incorrect in these "split" records.

Additionally, the test for INTRVTM=0 as criteria for the

first of a split record was changed to TSMSSEGN=1.

Thanks to Richard Morris, Progressive Companies, USA.
Change 08.260 Many variables in the HSM user SMF record accumulate from

DIFFHSM record to record. The new DIFFHSM provides the logic to

TYPEHSM de-accumulate these ascending values with the SAS DIF()

VMACHSM function. For the first occurrence of a "BY Group" value

Mar 25, 1991 or if values between adjacent observations decrease, MXG

sets the DIF()'ed variable to a missing value. This new

algorithm has been tested, but only with a few sets of

HSM SMF data records. If you have added HSM SMF record

processing to BUILDPDB, you must also %INCLUDE the new

member DIFFHSM in the EXPDBOUT member. If instead you

use member TYPEHSM to process HSM, the DIFFHSM member is

now already included for you. VMACHSM was also changed.

Variables DSRFUNCT,FSRFUNCT,VSRFUNCT were redundant with

DSRTYPE,FSRTYPE,VSRTYPE (which also take less space and

are decoded by an MXG format) so the ...FUNCT variables

were removed from VMACHSM datasets. Several PD4. variable

input statements were protected for hex nulls, and indent

of code was standardized. Some inconsistent data values

were found in IBM records - for example, several of the

fields which contain "number of tracks" contain hex

FFFFFFnn values, which is 4,294,000,000+ in decimal!

Use with caution, and let me know if you find IBM APARs

are needed to correct their data errors!
Change 08.259 TYPE6156 variables VOLSER1-VOLSER5 can be wrong if there

VMAC6156 are more than five volumes, because these variables were

Mar 24, 1991 not initialized. Line 020700 was change to a DO group:

IF NR6XVOLS=1 THEN DO; VOLSER1=VOLSER; VOLSER2=' ';

VOLSER3=' ';VOLSER4=' ';VOLSER5=' ';END;

Thanks to Kenneth D. Jones, Maritime Telegraph and Telephone, CANADA.

Change 08.258 New dataset PDB.SPUNJOBS is created automatically in your

SPUNJOBS MVS PDB by the new SPUNJOBS program, which reads today's

Mar 15, 1991 SPIN datasets and combines those SPINning job's records,

giving you PDB.JOBS variable names and a single dataset,

PDB.SPUNJOBS, describing jobs still spinning. Of course,

many variables in PDB.SPUNJOBS will have missing values,

because only fields from records that were found will be

non-missing for a particular job. Variable INBITS shows

which records were found, and in section "PDB" of Chapter

40 of the MXG Supplement is described which variables in

PDB.JOBS and PDB.SPUNJOBS come from which record.
Change 08.257 Goal Systems Explore/VM records for VM/370 were supported

VMACVMXP by MXG's VMACVMXP, but the VM/XA extensions were not in

XMACVMXP that data. This change stores the original member in the

Mar 15, 1991 XMACVMXP member (for backup, but will eventually go away)

and replaces the contents of VMACVMXP with new code that

was contributed, but only syntax checked. Use with the

appropriate caution.
Change 08.256 IMS Log Processing has been once again been significantly

TYPEIMS re-engineered, and now supports WFI (Wait-for-Input) and

VMACIMS multi-transactions per program schedule correctly, and it

Mar 14, 1991 also eliminates those occasional negative values in the

input and output queue times. The new design uses MXG's

BUILDPDB logic to merge the PRG and INOUT3 (message) data

and puts non-matches in UNMATCHD, but also creates a new

dataset MSGMISSD if the log does not contain all messages

for a program sked, (if WFI runs from 8am to 8pm, but if

the IMS log starts at noon, messages executed before noon

will be counted in the 07 record when the WFI terminates,

but would not have their 01-03-31-35-36 records in this

IMS log dataset). The CPU and DL/I calls for those missed

messages will be lumped into one observation in MSGMISSD

(variable NMSGMISS counts the missed message executions)

and all IMS resources are captured in IMSTRAN + MSGMISSD,

but only IMSTRAN captures transaction response times too.

The re-design eliminated the need to 'explode" IMS07 obs.

This redesign answers all reported problems with IMS log

processing for non-fastpath transactions, but at best the

IMS log is a poor source of IMS measurement data. The IMS

log was designed ONLY for backup/recovery, and IBM does

not care enough for IMS users to enhance the log to put

a transaction identifier in each record, making heuristic

algorithms based on found-data-patterns of MSGDRRN and

DEST necessary to reconstruct a transaction response from

its log records). Even when reconstructed, the IMS log

contains only a single measure of CPU time used which not

only includes the Control Region and Message Region times

but also includes the CPU time used when IMS transactions

access DB2! Large IMS sites deserve better quality from

IBM. Nevertheless, we will continue to do our best to

extract whatever can be captured from IMS log records.
Change 08.255 IMACJBCK allows SMF record selection by jobname, and this

IMACJBCK is a documentation note rather than a change. Only jobs

Mar 13, 1991 with blank or higher jobname are kept by MXG's default.

Type 30 records with nulls are created by IBM errors (see

MXG Technical Notes in Newsletter NINETEEN), and these

"bad" records were deleted by MXG because of the IMACJBCK

default. The default was needed to protect the PDB from

a runaway software monitor that created scores of records

with invalid job names, and the default could probably be

changed (to pass any or all values of jobname) without

any significant impact, but since that can't be proven in

advance, it seemed better to leave the default alone, but

remind you of this MXG default here and in comments in

the IMACJBCK member.

Thanks to Mike Revelette, Defense CEETA, USA.
Change 08.254 Support for OPPSI Operator Single System Image in SYSPLEX

VMAC76 environment added six new traceable variables to TYPE76:

Mar 6, 1991

OMDGWTOI - Lines of messages, WTOs, issued per millisec

OMDGCMDI - Commands issued per millisecond.

OMDGWTLI - WTLs (write to logs) per millisecond.

OMDGWQEB - Max WQEs (WTO Queue Elements) on queue.

OMDGOREB - Max OREs (Operator Reply Entries on queue.

OMDGAMRE - MAX messages on AMRF (Action Msg. Retention)
You must first enable RMF trace for these fields, and use

a PRINT of TYPE76 dataset to examine their values.

Change 08.253 SYNCSORT SCZ 33038 for SYNCSORT 3.3 or 3.4 adds two new

VMACSYNC counters with the number of frames of Hiperspace/Data

Mar 5, 1991 Space allocated (HPALLOC) and used (HPUSED). The two new

fields occupy formerly reserved fields, so prior versions

of MXG will not fail when this SCZ is installed.

Thanks to Sam Sheinfeld, Kraft General Foods, USA.

Change 08.252 Talk about great response from IBM for this vendor!!!

VMACDCOL At SHARE 76 last week, IBM mentioned that a future APAR,

Mar 5, 1991 OY37378, would add many new fields to the HSM DCOLLECT

record. Starting with the HSM Hotline telephone number

(listed in Pat Kearney's GC38-7014), it took only three

short telephone calls and less than three elapsed hours

before my fax machine coughed out the DSECT of the new

HSM record produced by DCOLLECT! Due to this great

response from IBM, APAR OY37378 is supported in this MXG

version (transparently), and you won't need to install a

new version later this spring! The new variables are:
ORGEXPDT='EXPIRATION*DATE'

UMALLSP ='ORIGINAL*ALLOCATE*SPACE'

UMBKLNG ='BLOCK*LENGTH'

UMCREDT ='CREATION*DATE'

UMEXPDT ='EXPIRATION*DATE'

UMFLAG2 ='INFORMATION*FLAG*#2'

UMGDS ='GDG?'

UMLBKDT ='LAST*BACKUP*DATE'

UMLRFDT ='LAST*REFERENCED*DATE'

UMNMIG ='NUMBER OF*TIMES*MIGRATED'

UMPDSE ='PDSE?'

UMRACFD ='RACF*INDICATED?'

UMRECOR ='VSAM*RECORD*ORGANIZATION'

UMRECOR ='VSAM*RECORD*ORGANIZATION'

UMRECRD ='RECORD*FORMAT*BYTE'

UMRECSP ='RECALL*SPACE*ESTIMATE'

UMRELBK ='RELBLK?'

UMSMSM ='SMS*MANAGED?'

UMUSESP ='ORIGINAL*SPACE*USED'

UMCHIND ='CHANGED*SINCE LAST*BACKUP?'

UMDSORG ='DATASET*DSORG*ORGANIZATION'
These variables will exist, but will have missing values

or blanks until the PTF for the APAR is installed. See

also the discussion in Change 8.130 for DCOLLECT info.
The PTF for OY37378 requires additional prerequisites.

APARs OY41039 and OY41256 must be installed before 37378.

Change 08.251 This is a revision of and enhancement to Change 8.182.

BUILDPDB The CICS 3.1+ Statistics records are combined to create

BUILDPD3 four interval data sets CICINTRV, CICEODRV, CICINTRV, &

CICINTRV CICUSSRV. CICS Stats records are created at an INTerval,

Mar 5, 1991 (CICINTRV), at EOD (End of Day, or Shutdown), at REQuest