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

See Change 18.313.
Change 18.279 NPM Type 28 Subtype 'DC'x (VTAM CSM Buffer Pool data) had

VMAC28 INPUT STATEMENT EXCEEDED INPUT RECORD error because I

Nov 24, 2000 mis-read the documentation. The four fields input at the

Nov 27, 2000 end of the sub-subtype VCSVDTYP=1 record (VCSMAXFS thru

VCSCUREC) exist in the VCSDTYPE=2 record, where they are

properly input. The four lines in the VCSVDTYP=1 INPUT

statement were replaced by +4 (to skip over the four

undocumented bytes at the end of the 176 byte segment).

Nov 27: Variable NRTDTYPE now kept with other NRT vars.

Thanks to Bruno Peeters, Dexia Bank Belgium, BELGIUM.

Change 18.278 -Summarization of CICS Statistics into CICINTRV for "EOD"

ADOCCICI (sum all "INT" and "End-of-Day" shutdown records into one

VMXGCICI total observation for each execution of a CICS region)

Nov 23, 2000 did not work properly: INTERVAL=EOD was never defined in

VMXGSUM/VMXGDUR, causing missing COLLTIME and STARTIME,

and multiple executions were lumped together. Since EOD

is defined only for CICS "EOD" in CICINTRV (set by adding

MACRO _CICINTV EOD % in your IMACKEEP or IMACCICS),

VMXGCICI was revised if EOD was requested and now passes

INTERVAL=MYTIME, and MYTIME= DATETIME=COLLTIME;, values.

The end result is that you get one observation for each

APPLID COLLTIME, where COLLTIME=CICSSTCK, logically the

the READTIME of each CICS region execution.

Note that if you dump your SMF data at midnight, and

stop/start the CICS regions daily at, say 4am,

requesting EOD will create two observations in todays

PDB.CICINTRV, one with yesterday's CICSSTCK time,

for the interval from midnight until 4am, and one with

today's CICSSTCK, summarizing the interval from 4am

until midnight, and neither observation would have the

total resources for each execution.

-Additional enhancements were also made. The PDB.CICINTRV

dataset is now created with the variables in alphabetic

order (with SYSPLEX SYSTEM SMFPSSPN COLLTIME etc. first)

by inserting a LABEL statement in the ORDER= argument of

the final VMXGSUM invocation.

-Two variables that had been summed are now maxed from the

SMD data: MAX=SMDHWMPS SMDIFREE

-Two variables that had been summed are now maxed from the

SMT data: MAX=SMTHWMPS SMTNTASK

-One variables that had been overlooked is now summed in

the XMC data: SUM=XMCPWQ

-These variables that had been overlooked are now summed

in the XMR data:

SUM= XMRAMISM XMRFAIT XMRFANW XMRFAOPN XMRFAOT

XMRFATX XMRITOV XMRIWAIT XMRRC

-Member ADOCCICI now documents which variables from which

Statistics datasets are SUM, MIN, MAX, and MEANed, and

how interval datasets are sorted. All except the MEANed

variables are defendable: for some datasets with multiple

records per interval (LSRPOOL, FCR), I calculate average

values that are of dubious value, so disregard them if

they are useless. Sometimes, averages are useful for

trending, and in all these cases, you still can go back

to the original detail to examine specifics.

Thanks to Normand Poitras, ISM, Canada.

Thanks to Chuck Hopf, MBNA, USA.
Change 18.277 Cosmetic. Labels for R744SSTA, R744STAC, and R744STRC

VMAC74 were corrected from the original documentation and now

Nov 20, 2000 match the SMF manual's description.

Thanks to Raimo Korhonen, CompMeas Consulting Oy, FINLAND

Change 18.276 CICS TS 1.3 SAP Journal Format SMF records were output in

VMAC110 dataset CICSJOUR instead of CICSSAP. The MXG test

Nov 16, 2000 IF JCRUTRID='SA' AND JCRLL GE 250 THEN DO;

was false because JCRLL is not created for the GLRHTYPE=2

sub-subtype journal format record. Instead of testing

JCRLL, the variable LENUDAT is used to verify that there

are at least 250 bytes of journal data left to read:

IF JCRUTRID='SA' AND LENUDAT GE 250 THEN DO;

and also the test for 00D1 was similarly changed to:

ELSE IF JCRUTRID='00D1'X AND LENUDAT EQ 50 THEN DO;

Thanks to L. Theodorides, Alte-Leipziger, GERMANY.
Change 18.275 Support for Vital Signs VisionNet VSAM file.

TYPEVITA Work in progress; only the subtype '38'x has been decoded

Nov 16, 2000 and code is not yet fully structured. But it works.

Thanks to Craig Collins, State of Wisconsin IT Services, USA.

Change 18.274 JES3 only. A short SMF type 6 record (90 bytes) had a

VMAC6 short "I/O Data Section", with SMF6LN1=30, which MXG had

Nov 16, 2000 not ever seen before. Normally, SMF6LN1=52 because all

thirteen fields existed, and MXG read them all. This

record ended with field SMF6DFE='01C9'x, which itself is

also not expected. Now, MXG tests for SMF6LN1=52 before

inputting SMF6DFE and the four subsequent fields.

Thanks to Robert Sample, Atlanta Journal-Constitution, USA.

Change 18.273 MXG 18.09 only. NPM APAR OW37743 inserted a four byte

VMAC28 field in the CSL segment in the NPMSUBTY='A0'x ODLC SMF

Nov 16, 2000 type 28 record, and MXG attempted to support that APAR in

Change 18.254, using IBM documentation only and no test

data, but now I find out IBM fibbed. MXG coded to test

the bit that IBM said would be on when the new field was

present, but data now shows the bit is on in records that

don't have the field (and without the APAR!), causing an

INPUT STATEMENT EXCEEDED RECORD LENGTH error. MXG now

uses the LENAOF segment length to determine if the field

is or is not inserted in your record. The original code:

IF LCSLCON6='1.......'B THEN INPUT LCSLT3S &PIB.4. @;

IF LENAOF GE 192 THEN DO;

INPUT

was replaced by these statements:

IF LENAOF GE 192 THEN DO;

IF LENAOF GE 196 THEN INPUT LCSLT3S &PIB.4. @;

INPUT

Thanks to Richard Rich, California Stephen P. Teal Data Center, USA.
Change 18.272 This example that analyzes allocation time components

ANALALOC (including HSM recall) had "IOPDB.xxx", but that was

Nov 13, 2000 changed to "PDB.xxx" to be consisted with other ANALs.

Thanks to Forrest Nielson, State of Utah, USA.

Change 18.271 Variable STC14VSZ now is formatted MGBYTES. (so that

VMACSTC it prints pretty!).

Nov 8, 2000

Thanks to Chuck Hopf, MBNA, USA.

Change 18.270 The SARRU33 records from CA-VIEW for subtype 33 (DELETE

VMACSARR REPORT) reserved field before SV33LNES should be +2 and

Nov 8, 2000 not +1. This caused subsequent fields in the dataset

to be incorrect.

Thanks to Craig Raridon, RadioShack, USA.
Change 18.269 ASTEX records were not read under ASCII SAS; the test for

VMACDMON DMONREC='F1'x, 'F2'x, and 'F3'x must be changed to '1',

Nov 8, 2000 '2', and '3' respectively, as DMONREC was INPUT with

$EBCDIC1, it became an ASCII character value under ASCII

SAS.

Thanks to Neil Ervin, Charles Schwab, USA.

Change 18.268 DSA size variables (SMSDSALI,SMDSATO,SMSHWMDS,SMSHWMDT,

VMXGCICI others) were incorrectly summarized into CICINTRV.

Nov 5, 2000 Some size variables were summed when they should not be.

Nov 12, 2000 These SMSDSANM count variables are summed in CICINTRV:

SUM=SMSASR SMSCREL SMSCRISS SMSCSS SMSCSSCM

SMSCSUBP SMSDSR SMSEXTS SMSEXTSA SMSEXTSR

SMSFMREQ SMSGMREQ SMSHWMSS SMSPWWS SMSSOS

SMSSV SMSTSOS SMSUCSS ,

and these SMSDSANM size variables are maxed in CICINTRV:

MAX=SMSUSSCU SMSUSSHW SMSCSSHW SMSDSALI SMSEDSAL SMSDSAT

SMSEDSAT SMSHWMDT SMSHWMED SMSNPAGP

Sizes (current, hwm, total, cushion) of each specific

DSA (CDSA,RDSA,SDSA,UDSA,ECDSA,ERDSA,ESDSA,EUDSA)

were correct in the CICSMDSA detail dataset, and it

must be used for analysis by SMSDSANM (DSA name).

Also, ANALCISH produced correct values.

Thanks to William Sherriff, IBM Global Services, USA.
Change 18.267 Variables SMF14CDL, SMF14CDS, SMF14UDL, and SMF14UDS are

VMAC1415 now formatted with MGBYTES, and they contain byte counts.

Nov 4, 2000

Thanks to Chuck Hopf, MBNA, USA.

Change 18.266 Support for NETVIEW "Hardware Monitor" type 37 SMF record

VMAC37 APAR OW45728 added new variable BRFRIBID, the

Nov 3, 2000 Ring Number (for token-ring LAN) or the bus number (for a

CSMA or token-bus LAN) to dataset TYPE37.

Change 18.265 Corrections based on source code scanning.

VMAC90A -Variable NEWFWKLD was not in keep in VMAC90A.

VMACTMDC -Variable CNSRRBA was not created nor kept in VMACTMDC.

VMACOMVT -Variables ON29UNK1-OM29UNK5 were removed from KEEP list

VMACPMTR as they are now known, input, with correct names.

Nov 3, 2000 -Dataset OMVTTCPC was incorrect, because "NEW" string had

been left in columns 1-3, but now have been removed.

-There was no %%INCLUDE SOURCLIB(IMACZDAT) in VMACPMTR,

so variable ZDATE was never created in dataset PERFMETR.

Thanks to Freddie Arie, TXU Gas, USA.

We took real vacation in Chena Hot Springs, near Fairbanks, Alaska. No

email for six days, saw the aurora three nights, made 9,000 ham radio

contacts (at KL7RA, with six other hams, during the annual 48-hour CQ

World Wide DX contest: made 8,600 contacts during the first 24 hours,

and then aurora hit Saturday night, and we made only 400 more QSO's in

the last 24 hours of the contest!).

======Changes thru 18.264 were in MXG 18.09 dated Oct 24, 2000======
Change 18.264 Changes to this SAS/Graph report example for RMFINTRV and

GRAFWORK TRNDRMFI Workload reporting now exploits HTML and SAS V8.

GRAFSAMP -PDBOUT added, allows you to write the output graphics

Oct 23, 2000 catalog to a different location than the input PDB.

This is important if you are using SAS/CONNECT to read

the input data from a mainframe PDB but want to store

the output on your PC or LAN. The default is "PDB".

-New parameter HTMLPATH= (default is null) should point

to a directory where all of the JPG files and HTML code

created by GRAFWORK will be stored.

-Member GRAFSAMP is a sample execution of GRAFWORK based

on a mainframe PDB as input using SAS/CONNECT with an

output PDB on a LAN and HTML being generated.

Thanks to Chuck Hopf, MBNA, USA.

Change 18.263 Changes to this SAS/Graph report example for RMFINTRV and

GRAFRMFI TRNDRMFI reporting has been enhanced to exploit HTML and

Oct 23, 2000 SAS V8, and to remove the now-obsolete VMXGGOPT macro.

-DEVICE= operand now directly used the GOPTIONS statement

rather than using VMXGGOPT.

-New parameter HTMLPATH= (default is null) should point

to a directory where all of the JPG files and HTML code

created by GRAFRMFI will be stored. At the conclusion

of GRAFRMFI, point your browser at the RMFIFRAM.HTML

dataset in the HTMLPATH= directory to view the generated

WEB pages.

Thanks to Chuck Hopf, MBNA, USA.

Change 18.262 Changes to this SAS/Graph report example for tape usage

GRAFTAPE has been enhanced for usability, and new options that

Oct 23, 2000 exploit ODS and SAS Version 8 for HTML output. This

example uses data from MXG's Tape Mount Monitor and from

STK's HSC and LSM records, expecting these datasets in

your TREND library: TAPEMNTS TRNDTALO TRNDHSC TRNDLMS.

-New parameter HTMLPATH= (default is null) should point

to a directory where all of the JPG files and HTML code

created by GRAFTAPE will be stored.

-DEVICE=JPEG provides a good choice of devices to make

your output more accessible and useable.

-DISPLAY=NODISPLAY now has no effect, since with SAS V8,

no output is produced if this option is in effect.

-GOUTTYPE=INDEPENDENT was removed to eliminate a warning

that had no meaning, and NOLIST added to all executions

of PROC DATASETS.

Thanks to Chuck Hopf, MBNA, USA.
Change 18.261 TYPE74 data for PAV Volumes had percentages over 100% for

VMAC74 PCTDVACT, PCTDVCON, PCTDVPND, and PCTDVUSE, because MXG

Oct 21, 2000 did not average those values across the number of UCBs

that were used during that interval. But now, to match

the IBM RMF report philosophy, all of the duration-based

device percentages are now divided by NREXPOSR to give

the average percentage. Note that only the percentage

values are divided by NREXPOSR; the duration variables

are unchanged, so in a 15 minute you could record true

DEVACTTM=39:58, DEVCONTM=21:33, DEVPNDTM=15:45 (with

DLYDEVTM=04:53 and DLYOTHTM=10:52).
This table maps the MXG variable names for the TYPE74

measures of duration, milliseconds per SIO, and percent

of the interval duration to their IBM DASD Report tag:
IBM MXG Duration Per-SSCH Percentage
---- ACT DEVACTTM dne PCTDVACT

CONN CON DEVCONTM AVGCONMS PCTDVCON

DISC DIS DEVDISTM AVGDISMS PCTDVDIS

IOSQ IOQ DEVIOQTM AVGIOQMS dne

PEND PND DEVPNDTM AVGPNDMS PCTDVPND

CUB CUB DLYCUBTM AVGPNCUB PCTPNCUB

DB DEV DLYDEVTM AVGPNDEV PCTPNDEV

DPB DIR DLYDIRTM AVGPNDIR PCTPNDIR

--- OTH DLYOTHTM dne PCTPNOTH

RESP RSP dne AVGRSPMS dne

USE USE dne dne PCTDVUSE
where dne = "does not exist" = is not created.
Schematically, these fields are related:
|----------------------ACT--------------------|

|----------PND----------|--DIS--|-----CON-----|

|-CUB-|-DEV-|-DIR-|-OTH-|
MXG creates the "Other" Pending time because the sum of

individual pends (CUB + DEV + DIR) is less than the total

device PND time. IBM does not directly report that other

component of pending time (although RMF shows it if you

compare DPB+CUB+DB DLYs with PEND). This PCTPNOTH has

usually zero, but now with PAV volumes, it has been seen

to be quite significant in this DURATM=900 sec interval:
One PAV Volume. NREXPOSR=4. DURATM=900 seconds.

|---------------------2398--------------------|

|----------------------ACT--------------------|

|----------945----------|--159--|----1293-----|

|----------PND----------|--DIS--|-----CON-----|

|-0.0-|-293-|-0.0-|-652-|

CUB DEV DIR OTH
The "Other" Pend time of 652 seconds is significant, and

according to none other than Dr H.P. Artis, for PAV, that

Other Pend, PCTPNOTH, includes the response time of the

subsystem to receive, verify, and acknowledge the first

CCW of the channel program. PEND time ends when the

subsystem (i.e., the logical volume) acknowledges the

first CCW. (Included in Other Pend is the percent of

time when I/O was pended due to conflicting Data Extents

for a track range, that is, overlapping writes and reads

waiting for a Data Extent area that is already being

written.

This Other Pend is the I/O delay to shared users of the

same dataset due to PAV parallelism, but without PAV,

that contending job could have had a much larger delay:

in the old days with MIM, the job would have run into

a DSENQ event, been put in hold and still be waiting in

the JES2 Held Job Queue until the enqued dataset was

freed by the first job, or now, with no limit on number

of batch initiators, the job is waiting in the DSENQTM

duration, initiated and holding an initiator, awaiting

allocation, held due to DSNAME ENQ due to DISP=OLD.
Note also that the Connect duration of 1293 seconds is

greater than the interval duration; PAV drops the point

of exclusive control to the track range in the data

extent for writes, and any number of simultaneous reads

can be concurrent even for one track as long as there is

not a write active for that track. I had trouble with

this until I realized that PAV must be behind cache so

parallel I/O to the channel occurs, and even for a cache

miss, the back end can have a 6:1 raid scheme which would

also permit six parallel I/Os between cache and raid.

Note that EMC's PAV's do not have any miss-parallelism,

due to their choice of a raid-1 back end.

A recent MXG-L posting reported an ADABAS volume with

an average of 37 UCB's assigned, showing PAV might be

a real salvation for that old database system!

Thanks to Diane Eppestine, Southwestern Bell, USA.

Change 18.260 REPORT=DEVC, columns %DEV RESV, %ANY ALLOC and %MT PEND

ANALRMFR have no values. AVG NUMBER ALLOC value is incorrect.

Oct 21, 2000 Line 4036

Change IF DEVCLASS NE 20X OR DEVCLASS NE 80X

To IF DEVCLASS NE 20X AND DEVCLASS NE 80X

Line 4056

Change IF DEVCLASS EQ 20X THEN PUT @1 STORGRUP @128" " @;

To IF DEVCLASS EQ 20X THEN PUT @1 STORGRUP @;

Change 18.259 Under SAS V8, you cannot re-define MACRO _BLD005 using

BUILDPDB %LET MACKEEP= to add PROC SORT and DATA steps before the

Oct 21, 2000 include of BUILD005, although it worked under V6. Instead

you can redefine MACRO _EPDBOUT (or use member EXPDBOUT)

to insert you code, and that is actually the exit that

should be used. In this case, the user wanted to sort

and create a PDB.TYPE30_5 dataset of today's records.

I will pursue the V8 problem with SAS when time permits.

Thanks to Simon Briggs, Allied Irish Bank, IRELAND.
Change 18.258 Variable CPCMODEL was added to dataset TYPE70PR by Change

VMAC7072 17.138, and is now also added to dataset TYPE70.

Oct 18, 2000

Thanks to Dr. Alexander Raeder, Karstat AG, GERMANY.

Thanks to Harmuth Beckmann, Karstat AG, GERMANY
Change 18.257 Tivoli NPM Type 28 Subtype 14X NRT record caused either a

VMAC28 NONZERO AOF NPMSUBTY=14 or UNEXPECTED NRPDTYPE=1 error.

Oct 14, 2000 Only NRPDTYPE=2 records (IBM Router) had been received,

so MXG forced this error to validate the Cisco NRPDTYPE=1

record, which turn out to be validly supported by MXG if

the DO group writing this message is deleted and the test

IF NRTDTYPE=2 THEN DO; is replaced with

IF NRTDTYPE IN (1,2) THEN DO; /*IBM/CISCO */

Thanks to Sandra Mischianti, Nicus Software, USA.

Thanks to Ken Patterson, Chemical Abstracts Service, USA.

Change 18.256 The MG073CD format for SMF73CPD (Channel Path Type) now

FORMATS decodes bits: 10X:OSA EXPRESS and 11X:OSA EXPRESS DIRECT

Oct 14, 2000 for those types of channels.

Thanks to Bob Falla, Clarica, CANADA.

Change 18.255 A semi-colon was missing after _ETY30TD invocation, so if

VMAC30 you tailored _ETY30TD and did not end your tailoring with

Oct 14, 2000 a semi-colon, ERROR 79-322 EXPECTING SEMICOLON occurred.

Thanks to Lawrence Jermyn, Fidelity Investments, USA.

Change 18.254 Support for NPM APAR OW37743 (INCOMPAT) SMF 28 for 3746

VMAC28 devices if NPM is collecting TIC3 (ODLC) link resources.

Oct 12, 2000 A four byte field, LCSLT3S='ACTIVE PU*COUNT*PER TIC3',

was inserted into the CSL segment, trashing some fields

in MXG dataset NPMLANOD dataset. The APAR text notes

that TIC3 data can be created by an NPM NETCOLL command

naming a generic resource, such as ODLCLNLK or ALLLINES,

which will match any ODLC (TIC3) link resource cause the

TIC3 data to be captured and sent to NPM.

Thanks to John Wilmot, Midland Bank, ENGLAND.

Change 18.253 The IBM sample report had two columns labeled STG THLD

ANAL88 which we replicated, but the second column is STG FULL,

Oct 11, 2000 and we've corrected our Logger replica report.

Thanks to Tom Elbert, Fortis, USA.

Change 18.252 Support for APAR OW44456 which creates the new SMF 120

record from the Component Broker element of WebSphere

IMAC120 Application Server Enterprise (EE) Edition. WebSphere

VMAC120 Application Server Standard Edition (SE), Advanced

TYPE120 Edition (AE), and Enterprise Edition (EE) is an IBM

TYPS120 product suite that runs on NT, AIX, Solaris, OS/390 and

VMXGINIT more. The SE version includes JAVA and the WebSphere

Oct 10, 2000 Application Server, but the EE version adds the Component

Broker element, and APAR OW44456 provides SMF type 120

records to record measurement of the Component Broker

that runs on OS/390.

Note: none of these members exist; no doc yet.

Change 18.251 Even after Change 17.348, this SAS/GRAPH example failed

GRAFLPAR if you didn't have SAS/GRAPH. The Graph-only SYMBOL1,

Oct 10, 2000 etc., statements were protected by the %IF .. %DO group,

but the following four lines were not deleted.

Thanks to John Allgire, Group Health, USA.
Change 18.250 ERROR.TYPE110.SUBTYPE 2. CICS STATISTICS RECORD STID=126

VMAC110 is an MXG error. The STILEN is 280, but MXG had miscoded

Oct 18, 2000 STILEN-STILEN-272; where it now has STILEN=STILEN-280;.

The error caused zero observations in CICCFS6D dataset.

STID=126 is new CFDT Coupling Facility statistics.

-Two new data fields were added to STID=127 record for

the CICFS7D CFDT Server Table Access dataset.

Thanks to Bernard Cadet, Michelin, FRANCE.

Change 18.249 Variables IMSGTEXT and OMSGTEXT, the first forty bytes of

IHDRIMS the message text, are decoded from the IMS 01 and 03 log

VMACIMS records. They have been useful in fraud investigation!

Oct 5, 2000 In addition, new exit IHDRIMS has been added, after the

IMS record has been read, but before decoding, so that

unwanted IMS records can be deleted.

Thanks to Frank Baird, Oklahoma Department of Human Services, USA.
Change 18.248 Unused Change number.
Change 18.247 Support for Neon System's Shadow Server V4.5 SMF record

EXSHDW01 creates three datasets:

EXSHDW02 dddddd dataset description

EXSHDW06 SHDW01 SHADOW01 SHADOW END CONNECTION

IMACSHDW SHDW02 SHADOW02 SHADOW INTERVAL

TYPESHDW SHDW06 SHADOW06 CLIENT REQUEST

TYPSSHDW The SHADOW01 observations may be end of Session, SM01RCTY

VMACSHDW ='S', or if the Logging feature is enabled, you will also

VMXGINIT get Interim Interval (='I') and Final Interval (='F').

Oct 4, 2000 Only subtype=1 data has been validated.

Thanks to Jim Gilbert, Sabre, USA.
Change 18.246 DB2 Report PMAUD03 failed (Uninit error for IF, NE, THEN)

ANALDB2R because the semicolon was missing from this line:

Oct 3, 2000 @90 'ORIGINAL AUTHID: ' QW0087OP;

Thanks to Bill Hamilton, Scottish Widows, SCOTLAND.

Change 18.245 Support for new NTSMF Objects:

EXNTADSM dddddd dataset Object

EXNTCFSE NTADSM ADSMCLNT ADSM Client Performance

EXNTMQQU NTCFSE COLDFUSE ColdFusion Server

EXNTWSCT NTMQQU MQQUEUES MQSeries Queues

EXNTWSQU NTWSCT WRLDMSCT WorldSecure/Mail Message Count

EXNTWSSM NTWSQU WRLDMSQU WorldSecure/Mail Message Queue

IMACNTSM NTWSSM WRLDSMTP WorldSecure SMTP Relay.

VMACNTSM The ColdFusion object still has errors in their data that

VMXGINIT Demand Technology is working to resolve. All except the

Oct 1, 2000 ADSM object's records have been tested.

-In addition, the "Full Image" object is now supported, as

it's counters are the same as the "Image" object with the

only difference being the instance name. In the Full