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

BUILDPD3 is no longer deleted by BUILDPDB, making it available for

Aug 10, 1988 your own use, if needed, following BUILDPDB logic.

Change 06.107 Correct a syntax error (period should be semicolon) and

WEEKBLD eliminate the non-existent PRENTIME variable in the BY

Aug 10, 1988 list for PDB.PRINT.

Thanks to Paul Bergman, Prudential, USA.

Change 06.106 VM/XA SP1 data from several sites has been used to update

VMACVMXA processing logic. The VXBYUSR user interval data is now

Aug 10, 1988 built from VXUSEACT and VXUSEINT data sets with logic to

recognize new sessions unambiguously eliminating negative

deaccumulations in the user data. (Negative delta values

were due to either insufficient delta logic or from two

byte counters which wrapped.) With additonal test data,

it has been possible to validate which fields really are

accumulated and additional variables are now correctly

DIF()'ed.

Thanks to Chuck Hopf, Dean Witter Reynolds, USA.

Thanks to Vic Fadool, Health Net Corporation, USA.

Thanks to Julie Roland, IBM Kingston, USA.

Change 06.105 MXG has been tested under SAS Release 5.18 and there are

EXITMONI three problems which induced MXG circumventions:

GRAFRMFI a.The EXITMONI member which assembles the TMON INFILE exit

IMACPDB (decompression algorithm for Landmark's CICS data) must

Aug 12, 1988 be re-assembled, using the SAS 5.18 macro libraries. The

internal filetable used by the SAS ASM macros increased

by 16 bytes. You cannot use the load members built with

5.16 macros under SAS 5.18, and the load members built

with the 5.18 macros cannot be used under SAS 5.16! You

can link the load modules into your SAS load library, or

keep two separate load libraries for the exit. The MXG

instructions for EXITMONI are on pages 59-60 of the MXG

Supplement, but only the BMACLIB library (file 22 on the

SAS distribution tape, unloaded by JCL in BTSOSRC in the

SAS JCL control library) is now required for the exit.

b.The KILL with PROTECT= option of PROC DATASETS no longer

works in SAS 5.18 (it was an un-documented feature - see

SAS User's Guide page 877.) Unless you have enabled the

PROTECT= feature (by removing comments in IMACPDB), SAS

5.18 has no problem. The change to PROC DATASETS removed

KILL and now requires a DELETE statement naming each data

set with PROTECT= to be deleted. Member IMACPDB now has

the necessary explicit PROC DATASETS code for those who

have chosen to protect their PDB data sets. In addition,

the (PROTECT=ZZZZZ) string has been externalized into a

new macro _PROTECT defined in IMACPDB so that you can go

either way easier.

If I had to do it over, I would never have used the

PROTECT= option in BUILDPDB. Protection was a holdover

from the days when SAS data sets were members of PDSs

and it helped avoid B37 abends and PDS dead space. If

I could I would make no-protect the default in MXG now

but that would require all existing sites to either

scratch and re-allocate each PDB data set or to run a

separate PROC DATASETS (from IMACPDB).set. So to avoid

the un-necessary work and exposure, PROTECT= will be

unchanged in future versions. PDB data sets now with

it still have it and no new protected data sets will

be created. Furthermore, IMACPDB lets you override any

way you want it. The following PDB data sets have been

and will continue to be built with PROTECT=ZZZZZ:

CICSACCT CICSEXCE CICSYSTM DDSTATS IPLS JOBS

PRINT SMFINTRV STEPS TAPES

TYPE70 TYPE71 TYPE72 TYPE73 TYPE73L TYPE73P

TYPE74 TYPE75 TYPE78 TYPE78CF TYPE78CU TYPE78IO

TYPE78PA TYPE78SP TYPE78VS

c.PROC GPLOT now sets an error condition if all values of a

variable are missing on a plot page. In 5.16 only a note

was printed. (Note this change was not made to the GCHART

CHART or PLOT procedures!) This has caused GRAFRMFI from

MXG 5.5 to ABEND if ERRORABEND is specified. SAS now has

a ZAP which restores GPLOT to its former self (the fix is

ZAP Z518.5694). The option NOERRORABEND has been added

to GRAFRMFI in case you have not installed that zap. In

addition, an unrelated, obscure logic error in the macro

logic in GRAFRMFI was uncovered and corrected.

d.GRAFRMFI exploited the GROUP statement so you could get

groups of graphs in batch (interactively, you simply

flag each graph to be grouped by PROC GREPLAY.) Changes

to the Graphic catalog in 5.18 and the lack of a way to

define "group" when building graphs has invalidated the

use of GROUP in GRAFRMFI.
---Changes through 6.104 completed the 7/27/88 2nd prerelease of 6.3--

Change 06.104 Pre-release VMXA code was not prepared to handle multiple

VMACVMXA "packets" of VM/XA monitor data from multiple monitor

Jul 26, 1988 sessions in one input file. Major changes were made to

the MXG logic to recognize the start of a new monitor

session in the de-accumulation logic. The data sets that

create the VMXAINTV consolidated interval data set have

been much more extensively validated, although there are

already twelve known problem areas under discussion with

IBM that I believe are problems in data (for example,

the Schedule records don't contain a processor address,

so it is impossible to validate them for guests which

allocate two or more virtual machines). VMXAINTV and the

sample data sets from which it is built seem to be okay.

However, VXIODDEV, VXSTOASP, VXUSEITE and all VXSCL...

data sets have occasional negative values as minimum,

and are under active investigation. Because VMXAINTV is

believed to be valid and useful, this tape was created.

Change 06.103 DB2 trace type 102 SMF record (new in 6.3) was not fully

ANALDB2R tested, causing FORMAT NOT FOUND errors.

FORMATS RENAME IMACDB2T to IMAC102.

IMACDB2T In member FORMATS change all occurrences of:

IMAC102 $MGDZDBC. to $MGD106C.

IMAC102A MGDZPID. to MGD106I.

IMAC102B In member VMAC102 change all occurrences of:

TYPE102 MG106I. to MGD106I.

TYPE102A In members ANALDB2R, IMAC102 (after RENAME), TYPE102 and

TYPE102B VMAC102 change all occurrences of IMACDB2T to IMAC102.

VMAC102 As documented in IMAC102, there are too many iterative DO

Jul 22, 1988 statements to build all T102.... data sets concurrently.

Members IMAC102A, IMAC102B, TYPE102A and TYPE102B create

half each and are used in JCLTEST and JCLUXREF for MXG

testing and to create DOCVER entries now.

Thanks to Scott Peterson, Southern California Edison, USA.

Change 06.102 ROSCOE 5.5 RTM response time data could be missing if the

ANALROSC SMF data was dumped after ROSCOE started. Change line 254

Jul 21, 1988 START=0; to read START=ROSTIME; to fix.

Thanks to Mike Doyle, Purdue University, USA.

Change 06.101 Member AAAAAAAA is a copy of member COPYRITE which will

AAAAAAAA be seen first if the MXG Source distribution library is

Jul 21, 1988 scanned (since that IEBUPDTE format is alphabetic). This

will simply make it clear what you've got.

Change 06.100 ACF2 sometimes contains ACTMFCBT data which is over 400

VMACACF2 bytes long, but MXG failed to protect, causing STOPOVER.

Jul 19, 1988 Find the line INPUT ACTMFCBX $VARYING200. ACTLTH @;

and change it to IF ACTLTH GT 0 THEN INPUT +ACTLTH @;

to skip over the excess data.

Thanks to Larry Cecil, First National Bank of Chicago, USA.

-------Changes through 6.099 completed the First pre-release of 6.3 -
Change 06.099 Bulk Data Transfer (BDT) variable SMF59NDT in line 184

VMAC59 is TODSTAMP8., not SMFSTAMP8 (although all other BDT

Jul 16, 1988 time stamps are SMF format!)

Thanks to Paul Walker, IBM, USA.

Change 06.098 Apparently a new release of NLDM has changed the size of

VMAC39 the accounting section. VERSNLDM was 000Dx, LENACCT=52.

Jul 16, 1988 Now VERSNLDM is 0014x, LENACCT=50, NLDMRSV1 removed!

On line 301 add @; after REVISION PIB2., replace 302:

IF LENACCT EQ 52 THEN INPUT NLDMRSV1 PIB2. @;

and add INPUT to beginning of line 303.

Thanks to Paul Walker, IBM, USA.
Change 06.097 NETSPY variables ARSPNET, T1RSPPC-T4RSPPC and OVRSPPC

VMACNSPY were sometimes "retained" into the next observation due

Jul 15, 1988 to lack of re-initialization. Now re-initialized.

Thanks to MP Welch, Sisters of Charity of the Incarnate Word, USA.

Change 06.096 Landmark's MONITOR data sets now have individual macro

EXITMONI names defined in IMACMONI which control the destination

EXMONDBD DDNAME to which the four data sets (MONISYST, MONIDBDS,

EXMONSYS MONITASK, and MONIUSER) are written, permitting the high

EXMONTSK volume data sets to be written directly to tape. The

EXMONUSR default DDNAMEs, WORK, were not changed. In the JCL for

IMACMONI the TMON exit installation (in EXITMONI), the DDNAME of

Jul 15, 1988 SYSGO should have been SYSLIN.

Thanks to Robert Manalo, EDS, USA.
Change 06.095 VM/XA SP1 data validation corrected several values and

VMACVMXA formats in the detail data sets. More important, this

Jul 12, 1988 change adds the VMXAINTV data set creation. This is the

consolidated interval data from the sample data from the

system, storage, processor and user domains. With about

228 variables per time interval, it will be the primary

source of VM/XA analysis, yet it holds 24 hours of half

hourly intervals in 1 track! Read comments in VMACVMXA.

Only simple sample reports are created now, but you can

expect more in the next pre-release. VMXAINTV should be

the primary data for investigation of VM/XA response,

resource utilization, etc. Note that the new response

measures provided are the average time and count of both

trivial and non-trivial interactions and are recorded in

two separate categories, UP and MP. CMS response will be

in the UP variables, while the MP variables contain the

response for MVS guests (or any other guest machines that

define more than one virtual CPU). Variables CALMP... and

CALUP... contain response information. CPU utilizations

are also calculated (PCT.....) to reflect overall CPU and

system versus user CPU utilization of all engines.
An unrealistically large test run (ALL possible records

from 60 intervals with 1500 active users from a 3090-600

created 100 MB of raw monitor data. MXG created all data

sets and all possible variables in 369 3090-600 CPU secs

and used 82 MB of 3380 DASD space. The VMXAINTV data set

was only 2 tracks! A smaller test sample (only 7 MB of

raw data) used 11 MB 3380 DASD and 55 3090 CPU seconds

to create all possible variables in all possible datasets

Change 06.094 NPMSLHTM should be input as PIB4.1 vice PIB4.2 to be

VMAC28 consistent with NPM 1.2 internal value, although this is

Jul 12, 1988 not documented by IBM and has not been fully confirmed.

Thanks to Henry Solomon, Merrill Lynch, USA.

Change 06.093 DOS variable LINES in keep list should have been LINEPRT.

TYPEDOS

Jul 12, 1988

Thanks to Jim Groseman, Duquesne Systems, USA.

Change 06.092 Support for NETSPY Release 3.1 added variables CCUNAME,

VMACNSPY LINENAME, and NETWADDR to the NSPYLU data set, along with

Jul 12, 1988 $MGNSPXD format addition and TRANSNO alteration based on

new flag bytes.

Change 06.091 Support for EPILOG 1000 (CICS) Version 4.30 added the GMT

IMACEPIL offset variable CIGMTOFF. Candle took our suggestion and

VMACEPIL created an INFILE exit named EPILOC that decompresses the

Jul 15, 1988 EPILOG CICS data. (Candle provides the documentation, the

ZAPs to SAS and the two load modules to be installed.)

Note that exit load moudules can be concatenated behind

STEPLIB or SASLIB libraries. After you install the exit,

you must then change the INFILE statement in MXG member

IMACEPIL (to add EPILOGC as described therein) to tell

MXG the exit exists.

Change 06.090 Still more undocumented treats from NPM 1.3. Line 1268

VMAC28 added test for NPDSUBTY EQ 041X to set AOFTYPE='NAD' to

Jul 11, 1988 read undocumented NAD(S) data and eliminate reams of

'AOF DNC-SESSION END' notes on MXG SAS log.

Thanks to Paula Lamphear, The Strohs Brewery, USA.
Change 06.089 NETSPY cluster and luname were not read for subtypes 10,

VMACNSPY 11, 20, 21, and 43 (hex). Line 412's ELSE IF statement

Jul 8, 1988 was expanded to add these additional five subtypes. Also

SECONDARY is now correctly spelled in line 179.

Thanks to Rodney L. Reisch, General Electric Silicon Products, USA.
Change 06.088 IMS 2.2 changes the location of the DEST field in IMS log

VMACIMS records type 35 and 36, as discovered by this MXG user.

Jun 29, 1988 The associated INPUTs are now conditional on the version

value (defined by you in member IMACIMS macro _IMSVERS).

Thanks to U. Guenschman, GEZ, Cologne, GERMANY.
Change 06.087 The MONITOR from Landmark accumulated CICS system data in

TYPEMONI MONISYST data set, which is deaccumulated into the actual

Jun 29, 1988 interval values by the ANALMON logic. This logic fails if

multiple CICS system's data was dumped (individually) and

then concatenated as input to MXG, because there is no

flag in the Landmark data to indicate a monitor startup.

This contributed change replaces the existing ANALMON and

heuristically detects a change in system to correctly

deaccumulate multiple-systems MONITOR data. (Note that

the original ANALMON macro still exists in TYPEMONI, but

the new macro (with the same name) is redefined after the

original definition. SAS uses the last macro definition

encountered.)

Thanks to John Leath, Fleet Information Inc, USA.

Change 06.086 The SORT order of the PDB.PRINT data set was incorrect in

MONTHBLD both MONTHBLD and WEEKBLD and in the PDB.PRINT section of

WEEKBLD Chapter 40 of the MXG Supplement. The actual order is:

Jun 29, 1988 READTIME JOB JESNR PRINTIME PRENTIME OUTDEVCE SYSTEM.

Only one run at one site actually had an error, but the

correct sort order is now used in these two members.

Thanks to Jim Bentley, US National Bank, USA.
Change 06.085 The IDMS Performance Monitor data has been enhanced and

TYPERTEJ further compared with the Cullinet reports. Their reports

VMACRTE are wrong. Their "Start Time" is actually the End of the

Jun 29, 1988 interval. Many report fields are 4 or 5 print positions;

Cullinet truncates the high order characters, which made

30652 Storage Gets (MXG value) print as only 652! Labels

for PGMOVLU, PGMOVLIU, and PGMOVLNU were changed after a

phone call to Cullinet corrected the documentation. The

storage and program pool sizes in their reports may be

wrong. the data (and hence MXG values) are pages, yet it

seems they have 4K pages and 1/2K pages in their reports.

However, not only Cullinet made mistakes. The label for

ARAFNAME is now FILE, not FIELD, and variable IHDRCVN,

the Central Version Number, was added to all IDMS data

sets (to separate different IDMS regions internally).

Finally, for IDMS journal (rather than SMF recording),

RTLHDRST should have been RT1HDRST in TYPERTEJ.

Thanks to Dick Whiting, Precision Castparts, USA.

Change 06.084 Offsets for the EXD data in the SAR SMF record (which is

IMACEXD either type 6 or a user record) were low by 2. Offsets

Jun 29, 1988 are now 91, 103, 111, 121.

Thanks to Cathy Hellums, U.S. Sprint, USA.

Change 06.083 Sample Turnaround report failed if job had not actually

ANALTURN terminated. IF SUBSTR(INBITS,3,1)='J'; test was added to

Jun 29, 1988 keep only jobs which had a 30 subtype 5 term record.

Thanks to Cathy Hellums, U.S. Sprint, USA.

Change 06.082 Sample NETSPY reports raised SORT order error. SMFTIME in

ANALNSPY the SORT BY statements should have been DATE, to agree

Jun 29, 1988 with the PROC PRINT BY statements.

Thanks to Rick M. Odegard, Kimberly-Clark, USA.

Change 06.081 Inclusion of STOPX37 record processing in BUILDPDB/3

VMACX37 caused a SAS 353 error about arrays if the _CDEX37 entry

Jun 28, 1988 in PDB exit EXPDBCDE was not the last _CDE.... entry.

The semi-colon at the end of the (KEEP= ...) statement in

VMACX37 should not have been there, was the culprit, and

no longer exists.

Thanks to Paula Lamphear, The Strohs Brewery, USA.
Change 06.080 Decoding of CICS variable UOWID into its actual time

TYPEMONI stamp variable UOWTIME (Change 6.016) was revised when

VMAC110 yet another undocumented timestamp format was found. The

Jun 28, 1988 non-fatal "INVALID DATA FOR INPUT FUNCTION" note and the

hex dump of the record resulted, setting UOWTIME missing.

Multiple Region Option (MRO) CICS environments should

find that the combination of NETSNAME UOWTIME is constant

for the same transaction across the CICS regions (TOR,

AOR, FOR) for the same "Unit of Work". While the MXG

UOWTIME is constant, the 8-byte UOWID variable will only

be constant in the first six bytes; the last two bytes of

UOWID increment at SYNC points. This revised algorithm

now checks times for valid format, rather than trying to

guess where the time stamp should be! The change applies

to both IBM CMF type 110 and Landmark's Monitor data.

Thanks to Daniel Engstrom, Continuum, USA.

Change 06.079 Continued DB2 enhancements in testing:

ANALDB2R 1. ANALDB2R enhanced with additional DB2PM report from 100.

IMACDB2T 2. Preliminary testing of SMF type 102 DB2 Trace record is

TYPE102 provided by this MAJOR contribution from this user. Many

VMAC102 new T102Snnn data sets are created by TYPE102, which uses

Jun 10, 1988 the macros defined in VMAC102 that are re-defined and

controlled (to select desired trace CLASS) in IMACDB2T.

Not all of the subtypes coded herein have been validated.

Thanks to Scott Peterson, Southern California Edison, USA.
Change 06.078 VM/XA validation continues.

VMACVMXA 1.SYSZONE is added to MRHDRTOD and all TODSTAMP8. variables

Jun 15, 1988 to convert GMT to local time in all MXG datetime stamps.

2.DM=3 RC=9 code SKIP+48 in line 2972 vice SKIP+40.

3.USEINT divide by HF counts protected for zero divide.

4.CALMPNTC respelled as CALMPNCT. (July 5, 1988).

Change 06.077 MVS 2.2 uses JFCRSV38 bit (JFCDSEQN, data set sequence

VMAC1415 number was coded on label parameter) which was a reserved

Jun 7, 1988 bit in the field decoded in to variable LABEL. This fix

tests bit values rather than hex byte value, so that the

variable LABEL is corrected.

Thanks to Howard R. Stenzel, Financial Data Products, Inc, USA.

Change 06.076 DB2 Version/Release QWHSRELN is PK1.1 vice PIB1., and no
VMACDB2 /16 operation (four places). QWHSRELN will now contain a

Jun 7, 1988 value of 1.2, 1.3, etc.

Thanks to Scott Peterson, Southern California Edison, USA.
Change 06.075 The user type 'C0' VM/Account record created for RSCS is

TYPEVM decoded into the new VMRSCS data set, due to this user

EXVMRSCS provided enhancement. Also, tape attach records with zero

Jun 7, 1988 connect time from CA's DYNAMCMS are no longer OUTPUT.

Thanks to Dick Whiting, Precision Castparts, USA.
Change 06.074 Several changes from one user:

FORMATS: 1.3240 should have been 3420 (VM Tape format)

TYPERTEJ: 1.SYSTEM $CHAR4. added to FORMAT statement.

2.After line 18, IF LGRTYPE NE 6 THEN DELETE: added.

Jun 7, 1988

Thanks to Dick Whiting, Precision Castparts, USA.

Change 06.073 Calculation of CU602LEN should have set values of 9, 10,

VMACVMON or 11. Cause last device to be lost from data set.

Jun 7, 1988

Thanks to Rachel Billington, British Telecom, ENGLAND.

Thanks to Dick Whiting, Precision Castparts, USA.
Change 06.072 More IMS cleanup:

ANALDB2R 1.IF SXNUM GT should be IF SXNUM GE.

EXIMSOUT 1.IMSTRPNO spelled IMSTAPNO, MSGTRAN spelled MSGTRANS and

IMSRECNO added to DROP= list.

JCLUXREF 1.Comments improved, VERS5 changed to VERS.

TYPEIMS 1.Move INCONT,LOGONID,MSGLEN and ORGENT initializations to

before their respective conditional assignment.

2.M014 should be M0114 at line 468 (made MSGSZIN wrong).

3.NODENAME added to RETAIN at line 766.

VMACIMS 1.COMPCODE is now a $CHAR4. formatted $HEX8. The last byte

was lost when it was stored as a four byte numeric. (SAS

needs one byte for exponent if the field is numeric).

Jun 7, 1988

Thanks to Pete Shepard, Ashland Oil, USA.

Change 06.071 NPM 1.3 type 28 record Stop Collection (Subtype=2) record

VMAC28 has a shorter PMC section, causing STOPOVER. At line 825,

Jun 7, 1988 conditionally INPUT NPMDATA thru NPMUPPER IF LENTOF GE 34

Thanks to Barry Pieper, FBS Info Services, USA.

Change 06.070 Now LRECL=32760 is specified for input SMF file. The old

VMACSMF value of 32756 solved a problem introduced with SAM/E and

Jun 7, 1988 documented at that time; somewhere in the intervening 10

years, IBM lifted the constraint and now the correct SMF

LRECL=32760 is required (LRECL=32756 causes a STOPOVER

condition if data is really LRECL=32760).

Thanks to J. Cary Jones, Philip Morris, USA.
Change 06.069 Variable TYPEIOML in data sets TYPE75 and TYPE77 now is

VMAC75 updated to recognize 309x processors; it is not used in

VMAC77 MXG and this change should have little visible impact.

Jun 7, 1988

Thanks to Richard Evans, Mervyn's, USA.
Change 06.068 RMF 3.5 support for PR/SM and other MVS 2.2 changes that

EXTY70PR related to expanded JESNR range to 99,999 were added.

VMAC26J2 1.Support for 99999 JESNR from JBID added to VMAC26J2 and

VMAC26J3 VMAC26J3 for JES2 and JES3 purge records.

VMAC30 2.Support for 99999 JESNR from JBID added to VMAC6 for PSF,

VMAC434 JES2, JES3, and EXTWTR type 6 records. The OUTPUT for the

VMAC6 EXTWTR was relocated to follow the creation of JESNR.

VMAC7072 3.Support for 99999 JESNR and revision of TYPETASK decode

VMAC71 (still $CHAR4 long, but fourth position will be blank for

VMAC74 99999 JESNR in all type 30 data sets.

Jun 2, 1988 4.New variable TERMNAME (Terminal Symbolic Name) created in

all type 30, type 4 and type 34 data sets.

5.New variable DSSIZHWM (Data Space Storage, highwater mark

used during the step) only in TYPE30_4 data set.

6.New documentation notes for TYPE30_4, PDB.STEPS, etc:

IOTMTOTL: For a DIV object, this includes the IO connect

time for READs, WRITEs and REREADs. This I/O .

connect time for DIV objects is only captured

for the address space, and not by DDNAME.

IOTMTODD: The connect time from each DDNAME does not