and new line
IF GEPARTLN GT 0 THEN GEPARMLN=GEPARTLN+(GEPARMNR-1)*2;
was added before GEOFFSET=GEOFFSET+GEPARMLN+6;
Thanks to Waldemar Rathfelder, ADP TAYLORIX GmbH, GERMANY.
Change 16.375 Fixes logic dealing with the length of the dataset names
VMXGSUM in the INDATA= that cause warning messages about invalid
Feb 16, 1999 arguments to SUBSTR function. Improved KEEP logic now
uses NODETAILS options on PROC CONTENTS to avoid reading
the entire dataset just to find out what variables exist.
Change 16.374 Trending member for the PDB.SMFINTRV dataset (which is
TRNDSMFI from Type 30 Subtype 2/3 records and is created as
Feb 16, 1999 dataset TYPE30_V, controlled in member IMACINTV.
Change 16.373 Confused by the syntax for the new %LET statements in
UTILBLDP MXG 16.16? This macro will build the SYSIN stream for
Feb 16, 1999 you to run BUILDPDB completely from SYSIN using the new
%LET architecture. You tell us the SPINCNT desired, any
time differences, any additional SMF records that you
want to add to the BUILDPDB process, and anything that
you want to suppress, and this utility builds the sysin
stream, complete with comments.
Thanks to Chuck Hopf, MBNA, USA.
Change 16.371A If you were not using ASUMUOW's normal default (use the
ASUMUOW earliest transaction record as the real transaction name)
Feb 16, 1999 then some of the variables that are used to identify the
Feb 18, 1999 transaction (such as LUNAME, USER, TERMINAL, SYSTEM,
SMFTIME) may not have come from that real transaction
record, but were erroneously picked up from the last
transaction record in the unit of work, which might have
come from the SPIN library and might have blank values.
This change treats these identifying variables the way
that TRANNAME is, so that they all come from the same
record, and corrects the logic for SPINning records.
Thanks to Kevin Marsh, AT&T Solutions EMEA, ENGLAND.
Change 16.371 Subtype 0 records from CICS/ESA 4.1 caused UTILCICS to
UTILCICS fail with INPUT STATEMENT EXCEEDED RECORD error. Replace
Feb 16, 1999 IF '02' LE OLDVERCI LE '03' THEN SUBTYPE=0;
IF SUBTYPE EQ 00 THEN DO;
with
IF '02' LE OLDVERCI LE '03' THEN DO;
SUBTYPE=0; /* CICS 1.6, 1.7, 2.1 FORMAT DATA */
The real culprit was the MXG logic error, not the record!
Thanks to Andre van de Riet, Hudson Williams Europe, THE NETHERLANDS.
Change 16.370 New variable PCTMETGO was not calculated correctly; the
VMAC7072 test should be IF R723CVAL*RTSMAP(M) GE R723CVAL THEN DO;
Feb 16, 1999 so we compare the bucket's response time (instead of the
percentage value for that bucket) to the response goal.
At present, counting the first six buckets would produce
the same value, as the 6th bucket is the 100% bucket, but
this algorithm should protect future changes.
Thanks to Bruce Widlund, Merrill Consultants, USA.
Change 16.369 MXG 16.06-16.10. Cosmetic. Names of some DROP variables
VMXGCICI were truncated (DIFTQTEM,DIFIOCRS,DIFIOCWF,DIFSTL,DIFST)
Feb 16, 1999 and so they were not dropped from temporary datasets, but
the MXG datasets were built without error.
Thanks to Alex Torben Nielsen, TeleDanmark EDB, DENMARK
Change 16.368 MXG 16.09-16.10. Macro PMVCICS had not been defined in
VMXGINIT VMXGINIT for the new CMRDETL support added in 16.09.
Feb 16, 1999
Thanks to Stephen Mesiano, Sharp Electronics Corp., USA.
==Changes thru 16.367 were in MXG NEWSLETTER THIRTY-FIVE Feb 20, 1999==
Change 16.367 MXG 16.06-16.10. Big CPU increase in BUILDPDB due to
VMXGCICI CICINTRV revisions introduced in MXG 16.06. MXG's use of
CICINTRV VMXGSUM's KEEPIN= option (which creates KEEP= statement
Feb 9, 1999 to keep only the needed variables) turns out to be very
CPU intensive in CICINTRV/VMXGCICI, even when there are
zero CICS observations! Replacing KEEPIN= option with
KEEPALL=YES removed the excessive cost of KEEPIN= logic.
In a single invocation of VMXGSUM when you need to keep
only a few variables from a large number of variables,
and when you have many observations (e.g., summing ten
variables from CICSTRAN, which was its purpose) using the
KEEPIN= option still makes sense, as it will save lots of
work space during the summarization. But CICINTRV, with
its many invocations of VMXGSUM (and a KEEPIN execution
for each argument of each VMXGSUM), and with essentially
all variables kept, so no space could be saved, the CPU
cost of KEEPIN was greater than the runtime without it!
Mac reported his BUILDPDB increased from 324 to 2650 CPU
seconds, a factor of eight! In my confirming tests, I
isolated the increase to CICINTRV and saw it jumped from
64 seconds to 446 seconds, about the same increase ratio!
That high cost (one build of the OPTVAR dataset took 53
seconds) is because we have to use SAS macro language to
parse arguments into macro variables and then convert to
variables in data steps so we can sort and remove dupes!
Note that VMXGSUM turns off SAS log messages, especially
for the data steps executed by the KEEPIN= algorithms,
and so the sum of "Step Used CPU" log messages will be
much smaller than the "Total CPU Used" value. For Mac,
his Total was 2650 seconds, but sum was only 380 seconds;
over 2270 seconds of KEEPIN= were not printed on the log!
Thanks to Mac Moss, Burlington Industries, Inc.
Change 16.366 NTSMF dataset PROCESOR variable DPCQUERT replaced DPCQUED
NTINTRV but the old variable name was still used in NTINTRV,
Feb 9, 1999 causing a missing value in NTINTRV.
Thanks to Chris Weston, SAS Institute ITSV Development, USA.
Change 16.365 Type 42 subtype 9 (dataset TYPE4237) was validated and
VMAC42 revised, and subtype 19 records showed wholesale revision
Feb 9, 1999 from their earlier documentation. Revised in 10Feb tape.
Thanks to Edward McCarthy, Health Insurance Commission, AUSTRALIA.
======Changes thru 16.364 were in MXG 16.10 dated Feb 8, 1999======
Change 16.364 QA test and MXG 16.04+ revision cleanup analysis:
DOC -TYPETMS5: PROC DELETE DATA=MGTMSVL; added after last use.
Feb 8, 1999 -SPUNJOBS: Variable DATETIME now formatted.
-MNTHDB2S: Variable STRTTIME now formatted.
-VMACTMDB: Variables D6ACBSC/D6ACESC now formatted.
-VMACVMXA: Variable DELTATM is now LENGTH 4 instead of 8.
-VMACEDGS: EDGSV: and EDGSU: added to dataset labels.
-VMACIMS/VMACIMSA: dddddd: added to dataset labels.
-TYPEMON8: dddddd: added to dataset labels.
-VMACOPC: OPC20: added to dataset label.
-VMAC28: L028IN7: corrected to 028IN7:.
-VMAC42: TY42P2: was missing colon in label.
-VMACMDF: Dataset Label was added.
-VMACHMF: Dataset Labels were added where missing.
-DIFFTPX: Dataset Labels was added.
-DIFFNTCP: Dataset Labels was added.
These products are not yet converted to the 16.04 design,
because they are second phase processes that don't read
raw data and don't really need the additional macro names
but they will eventually be revised to be consistent:
TRND70,TRND70PR,TRND71,TRND72,TRND27GO
TRNDDB2A,TRNDDB2S,TRNDDBDS,TRNDIDMS,TRNDRMFI,TRNDVMXA,
MNTH70,MNTH70PR,MNTH71,MNTH72,MNTHCICS,MNTH72GO MNTHDB2A
MNTHDB2S,MNTHDBDS,MNTHRMF,RMFINTRV.
VMXGHSM - does read HSM catalogs, but is standalone.
These products have not yet been 16.04 converted, and
won't be until someone asks for it, or until a change in
data requires an enhancement in these either standalone
or archaic programs:
TYPECRAY, TYPEFACO, TYPEPDL, TYPEVLFC, TYPEWWW, TYPEZARA
TYPE200, TYPETPNS, TYPERRTM.
Change 16.363 16.09 only, JCLTEST6 only. The new VMACMVCI for CMRDETL
JCLTEST6 still had RECFM=S370VB from PC testing, and MVS SAS does
VMACMVCI currently accepts only RECFM=VB, (although SAS has plans
Jan 21, 1999 for a fix to make S370VB an alias on MVS in Version 8?).
Change 16.362 The dataset name in MACRO _LHSMFUN should be HSMDSRFU.
VMACHSM The typo, HSMFSRFU, has never been a real dataset.
Feb 5, 1999 This 16.04 error has impact only if you used the _LHSMFUN
macro as the destination of your own PROC SORT, or if you
used the new _SHSMFUN macro; that created PDB.HSMFSRFU
instead of PDB.HSMDSRFU (but since it has all of the
variables, you can just rename it back to PDB.HSMDSRFU
without re-read of your SMF data).
Thanks to John McCray, Huntington Services Company, USA.
Change 16.361 Support for IDMS Version 14 Journal Records. Most IDMS
TYPEIDMJ sites write to SMF rather than a journal, and the SMF
VMACIDMJ records were fine, but the TYPEIDMJ support for journal
Feb 5, 1999 format had not been updated from IDMS Version 12! This
change only works with Version 14 data; I didn't take the
time to write it to work with both, as I don't think you
can be still running Version 12, but if you need it, I'll
enhance it for the old version too!
Thanks to David Thorn, Dow Jones, USA.
Change 16.360 -VMAC110: 16.09 only, new time variables in CICDS and in
IMACICOM CICINTRV did not have TIME12.2 format.
VMAC110 -IMACICOM: several MQxxxxTM duration variables had the
VMACDB2 same label as the MQxxxxCN count variables.
VMACTMDB -VMACDB2: new variables QXCOORNO/QXCRGTT/QXISORR/QXSTREOP
TYPETMON and QZZCBPNX/QZZCSKIP were missing because they had been
TYPEMONI added to the KEEP= list for DB2STAT0 instead of DB2STAT1.
Feb 5, 1999 -VMACTMDB: variable LMRKFLG1 was listed twice in FORMAT
statement; the last instance - $HEX2. - was removed as
it was overriding the $MGTMDGF decoding format.
-Archaic TYPETMON and TYPEMONI were upgraded to the 16.04
architecture, mostly for QA purposes than real need, as
both are archaic (replaced by TYPETMO2 and TYPEMON8).
Thanks to Chris Weston, SAS Institute ITSV Development, USA.
Change 16.359 Change 16.204 externalized tailoring for IMACACCT among
IMACACCT others, but the &MACACCT; statement was not added to
Feb 3, 1999 the end of the IMACACCT member until now.
Thanks to Paul Oliver, BHP, AUSTRALIA.
Change 16.358 TYPE39_6 variables DURATM, STARTIME, in the SCS segment
VMAC39 were originally only created in Netmaster records, but
Feb 3, 1999 became unneeded when IBM added ACTSTIME/ACTETIME to the
accounting segment. I never noticed, but now in NetView
for OS/390 Version 1 Release 1 type 39 records, IBM has
put several character fields where those old fields were
input, causing strange dates (like 2008) when MXG reads
those characters as TODSTAMP datetimes! The unneeded old
Netmaster variables are now set missing/blank, and the
new fields (Primary/Secondary CP NetworkID/Name and APPN
class of service name and APPN transport priority) were
added to each of the subtype 1-8 TYPE39 datasets.
Thanks to Tracy Blackstone, Kaiser Permanente, USA.
Change 16.357 Variable DDNAME was not kept in TYPETMNT, but no one had
VMACTMNT need of it. But now, in developing a new MXG tool that
Feb 3, 1999 decomposes all job delays, we find DDNAME is needed so we
can detect "parallel" tape mounts (when one tape DD has
multiple tape drives allocated) so we can not-include the
second-and-subsequent mount delays in the job delay total
(the reason you allocate two drives for one tape DD is so
that the rewind/mount time of one is in parallel with the
job's use of the other, to prevent any delay to the job).
Change 16.356 Support for Sterling's SOLVE NetMaster for TCP/IP SMF
EXTYSOLN creates one observation in dataset TYPESOLN for each
FORMATS record (which are identified as either interval or
IMACSOLN event), decoding as many of the up-to-12 fields that
TYPESOLN might be present. Variable ATTRID identifies what is
VMACSOLN measured in each record, and has values in mixed case
VMXGINIT (like "ifInPktsNUcast" and "ifInPktsDiscard") for each
Feb 3, 1999 metric. There are nearly 60 metrics available:
TCP/IP Stack: FTP/TELNET/Workload Connection activity
TCP/IP Node Monitor: SNMP counters/response/available
CISCO Monitor: Channel card load/errors/TN3270 use
Local Interface: Status and Thruput
The complete list and description is in SOLVE:Netmaster
Graphical Performance Manager User's Guide, Appendix C,
Attribute Descriptions.
Thanks to Wendy Wong, Sterling (Field Engineer), AUSTRALIA
Change 16.355 The DB2 Buffer Hit Ratios calculated in DB2STATS became
VMACDB2 negative because the TSPP field apparently should not be
Feb 2, 1999 included. The old equation TGET-(TSIO+TDPP+TLPP+TSPP)
had values of 175-(101+0+0+1621) so clearly TSPP is out
of place, and thus has been removed from the equations
for B1HITRAT/B2HITRAT/B3HITRAT/B4HITRAT and BPHITRAT.
I am now looking for an IBM update to their equations.
Change 16.354 There are additional fields captured in HPs MeasureWare
VMACMWSU for Sun that can now be output; some fields were removed
Feb 2, 1999 and some were reordered. The new fields are defined, but
you may need to compare the headings of the report with
the MXG input sequence to see that they are the same.
Thanks to Tim Crocker, NCCI, USA.
Change 16.353 The RMF WKLD Workload Activity Report was updated for
ANALRMFR OS/390 MODE=COMPAT, the WLMGL Report RPTORP=RCLASS
Feb 2, 1999 SCLASS SCPER has been updated for MODE=GOAL, the WGPER
Feb 20, 1999 BY lists now have BY SYSPLEX STARTIME ....
Thanks to Jiann-Shiun Huang, CIGNA, USA.
Change 16.352 Some report/summary members had not been revised for the
ANALARB 16.04+ architecture and still %INCLUDEd their product's
ANALDB2R IMACxxxx member instead of its VMACxxxx member. As long
ANALPDSM as the program ran in the same step that built its data
ASUMHPSU sets, there was no error, because the VMACxxxx member
ASUMHPUX would have already been %INCLUDEd, but standalone runs
Feb 2, 1999 caused SAS "180" syntax errors due to missing macros.
MXG no longer defines macros in the product IMACxxxxs,
but instead MXG defines its macros in the VMACxxxx's,
which themselves include the IMACxxxx if you have one.
any overrides of the MXG defaults.
But not all includes of IMACxxxx members were changed,
as there are non-product IMACxxxx members (IMACRMFI)
that are still validly included by these programs.
Thanks to Chuck Hopf, MBNA, USA.
Change 16.351 -Support for BETA93 3.1.3 Subtypes 1, 2, 3, 4, 20, 40, 41
FORMATS and 42 has been validated with single-record tests. In
VMACBETA one subtype 20, the READ DATE and TIME are invalid, which
Feb 2, 1999 means variable READTIME (which is necessary to match the
Beta records to other SMF records for the job accounting)
has a missing value, so the Beta records can't be matched
to their jobs other data! In the one test record:
valid LINP date invalid LJOB date
S20LINPT+S20LINPD S20LJOBT+S20LJOBD
time cyyddd time cyyddd
'0064B47A 0095001F' '00000000 0100000F'x
That record was written on 2Nov98; the 95 date shows the
document has been around for a while. This discovery
will be forwarded to the MXG site to forward to the Beta
vendor for response and correction. I will update this
MXG note when I hear back from the vendor, but since the
invalid date causes a hex dump of the record on the SAS
log, I have added "??" to suppress the dump and message.
Create the Beta datasets and use PROC MEANS N DATA=....
to see how many records have invalid data for READTIME.
-Existing format MGBETRT for variable BETAATT4 was updated
since it now has character rather than hex values.
Thanks to Wolfgang Prescher, Quelle, GERMANY.
Change 16.350 Support for WINDOWS NT 5.0 (INCOMPATIBLE) for NTSMF adds
EXNTAPLT new objects and new counters, but NT 5.0 appears to have
EXNTGWNW removed some important data:
EXNTHTIS =Changes to existing supported objects:
EXNTINSF -SYSTEM object removed all of the CPU busy fields; these
EXNTINSR variables will have missing value in the SYSTEM dataset:
EXNTJOB APCBYPAS DPCQUERT DPCRATE DPCBYPAS INTERUPT
EXNTJOBD PCTCPUTM PCTUSRTM PCTPRVTM PCTDPCTM PCTINTTM
EXNTMACF But those variables are also in the PROCESOR dataset
EXNTPRNQ (although per-CPU instead of per-interval), and NTINTRV
EXNTRASI logic sums those PROCESSOR variables into the interval
FORMATS CPU busy measures (by accident, because it happens that
IMACNTSM the MERGE statement lists SYSTEM before PROCESOR, and SAS
VMACNTSM always takes values from the last dataset when the same
VMXGINIT variable is in multiple datasets), so you should make
Feb 1, 1999 sure you are creating the Processor object so you can get
Feb 4, 1999 those most-important measures in PDB.NTINTRV dataset.
And if you were using dataset SYSTEM in your reporting
for those fields, you'll need to change to use PROCESOR
or PDB.NTINTRV for the interval utilizations in NT 5.0.
Two new variables, PROCESES and THREADS were added to the
SYSTEM object to count processes and threads.
-LOGLDISK object no longer has the physical disk name,
variable PDSKNAME, so you can't recognize which logical
disks are on which physical devices! Two new measures
PCTIDLE (percent idle time) and SPLTIORT (SPLIT I/Os per
second) were added to this object.
-PHYSDISK object has the same two new measures, PCTIDLE
and SPLTIORT that were added to LOGLDISK object.
=New objects, their "NTdddd" suffix and dataset name:
suffix dataset
AppleTalk NTAPLT APPLETLK
Gateway Service for Netware NTGWNW GTWYNETW
Http Indexing Service NTHTIS HTTPINSR
Indexing Service NTINSR INDEXSRV
Indexing Service Filter NTINSF INDSRVFL
Job Object NTJOB JOB
Job Object Details NTJOBD JOBDETLS
MacFile Server NTMACF MACFILE
Print Queue NTPRNQ PRINTQUE
RADIUS Server [IAS] NTRASI RADIUIAS
This support is based on Windows NT 5.0 Beta 2 discovery
records, and some records have been tested, but none of
new objects yet. NT 5.0 is to be renamed Windows 2000.
With 486 processors and NTSMF earlier than 2.2, an INPUT
STATEMENT EXCEEDED RECORD LENGTH on the 0.0 record is now
avoided; the CPUSPEED fields are now only input if your
NTSMF version is 2.2 or higher.
Change 16.349 The statement SQRTARG=(SSQELAP/TRANS-(RESPAVG**2);
ASUMCICS caused INVALID ARGUMENT TO SQRT function because it
ASUMCICX should have been IF TRANS GT 0 THEN SORTARG=(....
TRNDCICS Change 16.137 protected for negative square roots, but
TRND72 did not protect for zero divide with zero transactions.
Jan 30, 1999
Thanks to Normand Poitras, ISM, CANADA.
Change 16.348 The internal _PDB6 macro was enhanced, adding variables:
BUILD005 BINUSED BIN1BNCT BIN2BNCT BIN3BNCT BIN4BNCT
BUIL3005 BIN5BNCT BIN6BNCT BIN7BNCT BIN8BNCT BIN1USED
Jan 30, 1999 BIN2USED BIN3USED BIN4USED BIN5USED BIN6USED
BIN7USED BIN8USED
so they are kept in PDB.PRINT to measure bin activity.
Thanks to John A. Rivest, TDS-Computing Services, USA.
Change 16.347 Preliminary report merges datasets HSMFSRST and PDB.JOBS
ANALALOC to report how much of ALOCTM allocation delay was due to
Jan 30, 1999 HSM recall (and recall time spent queueing for an HSM
Feb 17, 1999 thread is separated from actual recall time), and how
much was real allocation delay. Further enhancements to
bring in other delay records (e.g., MXGTMNT allocation
and tape mount delay) are planned, to ultimately provide
measurement of each separable delay in a job's life.
Revised, Feb 17, 1999, but does not yet account for
overlap of mount times when parallel devices mounts are
used (i.e., multiple tape devices for one DD).
Thanks to Rebecca Cates, PKS Computer Services, USA.
Change 16.346 Thruput Manager SMF records with no token entries caused
VMACTPMX ERROR.VMACTPMX. TOKEN NOT FOUND, FIELD WAS NOT INPUT.
Jan 30, 1999 because MXG logic assumed there would always be tokens.
Change the DO UNTIL (NEXLOC=0); to now read
IF OFFXFE GT 0 THEN DO UNTIL (NEXTLOC=0);
and the record will be output.
Thanks to Joseph Montana, Trans Union Corporation, USA.
Change 16.345 Enhanced HSM summarization now separates the HSM delay
ASUMHSM into active and queued requests so you can track HSM's
TRNDHSM concurrent thread count, used and required. Triggered
Jan 30, 1999 by the new ANALALOC program, in Change 16.347, and now
used by the revised ANALHSM in Change 16.383.
Change 16.344 RMF Reports are enhanced:
ANALRMFR -Coupling Facility, Average CF utilization corrected.
FORMATS PCTCFBSY, if logical processors present, calculate
Jan 30, 1999 using logical processor effective.
-Updates PARTITION DATA REPORT for deactivated LPARs.
The values for counts in the MXG reports, will be
output as they are in the MXG/SAS database. (IBM
reports counts as 80K but ANALRMFR shows 80396).
Some values (standard deviation, delay time and /all
time), are still in development/validation.
-ADD Subchannel Activity Report.
-Format MGRMFF was revised to round percentages better.
Thanks to Scott Wiig, USBank, USA.
Change 16.343 Using JCLIMSLG/ASMIMSLG logic to read IMS log records for
VMACIMSA IMS FastPath transactions created zero observations in
Jan 30, 1999 IMSFASTP dataset, because the variable SUBCODE was not
kept in the IMS591,593,596,597, and IMS598 datasets.
The variable was added to the KEEP= list for those five
datasets (but Pete just used the five _KIMS59x macros
to add SUBCODE to circumvent the error in the interim).
Also, the PUT _ALL_'s that were unlimited will now only
print for the first five instances of unmatched fastpath.
Thanks to Pete Gain, SAS Software Pty, ENGLAND.
Change 16.342 New variable CECSER, the CPU SERIAL NUMBER OF THE CEC,
VMAC7072 (Central Electronic Complex) is created from CPUSER in
Jan 29, 1999 TYPE70 and TYPE70PR datasets, to identify the hardware
box in which this MVS system ran. The last two bytes of
CPU Serial (CPUSER) are converted to hex representation
and stored in the new 4-byte character variable CECSER,
so a CPUSER='03870F'X has CECSER='870F'.
Thanks to Dr. Alexander Raeder, Karstadt AG, GERMANY.
Change 16.341 IMACPDB documentation. No code was changed.
BUILD005 The new architecture eliminates tailoring member IMACPDB,
BUIL3005 but I failed to document how! Previously, you EDITed the
IMACPDB macros (_PDB6, _PDB26J2, etc.) in IMACPDB to add fields
ZMACPDB (not kept by default) from type 6, 25, 26, and 30 SMF
Jan 29, 1999 records, into the PDB.JOBS, PDB.STEPS, and PDB.PRINT
Dostları ilə paylaş: |