The Association of System

Performance Professionals

The Computer Measurement Group, commonly called CMG, is a not for profit, worldwide organization of data processing professionals committed to the
measurement and management of computer systems. CMG members are primarily concerned with performance evaluation of existing systems to maximize
performance (eg. response time, throughput, etc.) and with capacity management where planned enhancements to existing systems or the design of new
systems are evaluated to find the necessary resources required to provide adequate performance at a reasonable cost.

This paper was originally published in the Proceedings of the Computer Measurement Group’s 1988 International Conference.

For more information on CMG please visit http://www.cmg.org

Copyright Notice and License

Copyright 1988 by The Computer Measurement Group, Inc. All Rights Reserved. Published by The Computer Measurement Group, Inc. (CMG), a non-profit
Illinois membership corporation. Permission to reprint in whole or in any part may be granted for educational and scientific purposes upon written application to
the Editor, CMG Headquarters, 151 Fries Mill Road, Suite 104, Turnersville , NJ 08012.

BY DOWNLOADING THIS PUBLICATION, YOU ACKNOWLEDGE THAT YOU HAVE READ, UNDERSTOOD AND AGREE TO BE BOUND BY THE
FOLLOWING TERMS AND CONDITIONS:

License: CMG hereby grants you a nonexclusive, nontransferable right to download this publication from the CMG Web site for personal use on a single
computer owned, leased or otherwise controlled by you. In the event that the computer becomes dysfunctional, such that you are unable to access the
publication, you may transfer the publication to another single computer, provided that it is removed from the computer from which it is transferred and its use
on the replacement computer otherwise complies with the terms of this Copyright Notice and License.

Concurrent use on two or more computers or on a network is not allowed.

Copyright: No part of this publication or electronic file may be reproduced or transmitted in any form to anyone else, including transmittal by e-mail, by file
transfer protocol (FTP), or by being made part of a network-accessible system, without the prior written permission of CMG. You may not merge, adapt,
translate, modify, rent, lease, sell, sublicense, assign or otherwise transfer the publication, or remove any proprietary notice or label appearing on the
publication.

Disclaimer; Limitation of Liability: The ideas and concepts set forth in this publication are solely those of the respective authors, and not of CMG, and CMG
does not endorse, approve, guarantee or otherwise certify any such ideas or concepts in any application or usage. CMG assumes no responsibility or liability
in connection with the use or misuse of the publication or electronic file. CMG makes no warranty or representation that the electronic file will be free from
errors, viruses, worms or other elements or codes that manifest contaminating or destructive properties, and it expressly disclaims liability arising from such
errors, elements or codes.

General: CMG reserves the right to terminate this Agreement immediately upon discovery of violation of any of its terms.
 Learn the basics and latest aspects of IT Service Management at CMG's Annual Conference - www.cmg.org/conference

SYSTEM WORKLOAD MANAGEMENT USING PC-BASED GRAPHICS

RONALD T. WELF
VISA INTERNATIONAL

Buy the Latest Conference Proceedings and Find Latest Computer Performance Management 'How To' for All Platforms at www.cmg.org
ABSTRACT

Computing systems resources at Visa International include seven 3090-class and three 4300-
class processors located in three geographically dispersed data centers. Due to the
diversity of operating system and application software which runs on these machines,
workload characterization and measurement comprise a large part. of the systems planning
effort which occurs each year.
Join over 14,000 peers - subscribe to free CMG publication, MeasureIT(tm), at www.cmg.org/subscribe

After a brief overview of the annual capacity and system configuration planning cycle,
this presentation focuses on use of the PC-based spreadsheets and presentation graphics
both to develop monthly workload projections and to track actual usage data against these
projections. A graphics approach is used to facilitate obtaining specific workload
projections from software development department managers. In addition, production
workloads are analyzed by application system and projections madE! in a similar fashion.

The use of spreadsheets and graphics provides a very flexible and responsive mechanism for
aggregating and trending workload-related usage data. These data are obtained from
MVS/SMF and VM/ACCOUNT records. This approach makes use of a host-based system accounting
package to summarize these detailed data into a form suitable for manipulation on a PC
workstation. These data are then used to compare directly to the workload projections,
giving line managers visual feedback regarding the accuracy of their individual
projections.

1.0 INTRODUCTION This paper is organized into an introduction

and seven major sections. Sections 2.0 and
3.0 describe briefly the systems environment
During the past several years, data and planning precess at Visa International.
processing systems at Visa International have Section ·4.0 addresses workload
grown very rapidly to keep pace with customer characterization, while Section 5.0 covers
demand. The annual peak periods produce workload forecasting. In Section 6.0 a
online authorization traffic exceeding 300 method for tracking the actual vs. projected
messages per second and nightly clearing and system workloads is described Finally,
settlement volumes reaching 12 million drafts Sections 7.0 and 8.0 describe future plans
in a single day. In addition, competitive and present conc~usions respectively.
pressures have spawned new products and
applications, and intensified the focus on
becoming even more cost-efficient in 2 .0 SYSTEMS BNVIR01IMBNT AT VISA
delivering services to end-users.
Computer processing facilities at Visa
Such rapid growth and cost consciousness has International are housed in three,
resulted in the need for a more formal and geographically dispersed data centers: San
ongoing focus on capacity management Mateo, California; McLean, Virginia; and
functions with direct links to the annual London, England. An additional facility in
bUdgeting process. While the complete range Canada is planned. The San Mateo facility is
of such functions can be grouped into three the largest, containing separate processor
categories workload characterization, complexes for credit authorization, Debit/ATM
workload forecasting and performance services, clearing/settlement and batch
prediction -- this paper concentrates on the support applicat:ions. Additional systems
first two categories. The focus is on the provide support for systems development and
use of PC-based spreadsheet and presentation corporate services activities.
graphics tools to accomplish the following:
At the San Mateo facility, the credit
Provide insight to management regarding authorization and Debit/ATM applications each
the nature of processing workloads: execute under a separate Transac~ion
Processing Facility (TPF) operating system
Communicate in non-technical terms, how and are not discussed further. The
these workloads change over time; and clearing/settlement and other batch
applications comprise the "production"
Provide a mechanism to assist in the workloads and are hereafter referred to as
process of projecting future changes and such. These Production applications execute
additions to existing workloads. under the MVS/XA. (Version SP2. 2) operating
system on an IBM 3090-150 processor .

• 540 •

Find a CMG regional meeting near you at www.cmg.org/regions

 Learn the basics and latest aspects of IT Service Management at CMG's Annual Conference - www.cmg.org/conference

percent of the, effort required by the systems

The Development complex is the largest, planning staff, that the PC-based tools
driven by an IBM 3090-200. The systems described below are most helpful.
software environment on the L.,-,,,elopment CPU
consists of VM/HPO (Version 4.2) and MVS/XA
(version SP2.2), both running under VM/XA-SP The Assumptions Review step is essential. It
(Version 1). There are approximately 600

Buy the Latest Conference Proceedings and Find Latest Computer Performance Management 'How To' for All Platforms at www.cmg.org
is at this point that management interviews
PROFS users and 450 VM/CMS users supported by are complete and projected growth and new
the VM/HPO system, and 150 TSO users applications assumptions are summarized and
supported by the MVS/XA system. Roughly presented to senior systems management for
20,000 batch jobs are executed each month on concurrence. Without this step, much of the
the MVS/XA system. technical work which follows could be
invalid, bein9 based on erroneous planning
assumptions.
Systems supported at San Mateo require
Join over 14,000 peers - subscribe to free CMG publication, MeasureIT(tm), at www.cmg.org/subscribe

approximately 250 volumes of 3380 DASD,

di vided approximately evenly among the TPF, Once the assumptions are reviewed and agreed
Production and Development environments. upon, a two-part step begins. The first part
Fifteen communications processors drive the c?nsists of combining the existing workloads
networks for remote access to the online wlth the growth ~ssumptions and projections
applications. The three centers in McLean, for new appli.cations to create a aggregate
London, and Canada are configured similariy, systems capaclty plan. The capacity plan
but without the Debit/ATM and Development c?vers . all major in-center subsystems
applications. The processes described below (lnc!udlng DASD, tape, communications
apply to these facilities also. processors, channel switches and
workstations) ,:is well as CPU workloads.
3.0 SYSTEMS PLANNING AT VISA
using this capacity estimate as a guide, an
The formal systems planning process currently effort is then undertaken to develop an
followed at Visa is relatively new and is initial projected system configuration.
still evolving. The process, which During this step, CPU channel layouts are
culminates in the preparation of a System created/modified and DASD space requirements
Environment Plan (SEP), is depicted in Figure are con:rerted into equivalent real "boxes'·.
1, S.E.P. Methodology. The top half of the In addltion to the technical aspects of
diagram represents the initial data gathering channel layout, CPU size, and memory
and budget submission ("GUESS") phase. It is requirements, 1:he following are considered:
in this phase, which represents roughly 50
CPU and peripheral equipment lease
RESERR[H conunitments and;
- Interviews
CONFIGURATION
• Raw Data Known budget constraints and other
Han1ware
f-o - Otlle, ~Ia,," organizational factors.
Prellmnary Pla~ - Con$ul~1n9 S~rvlces
~I lies ~,';':S~ Consideration ·:)f these non-technical factors
/ cannot be over-stressed, as they can become

~
the most critical ones.
ASSUMPTIONS

-..
!luslh'
•
-
-
-

M'JO.
Soflware
TeChnol(lgl'
Proj'Kts
Operatlor>al COr\5.d

,0> r,Fv'
411';168

11FT '''/l';'ABI
GUESS
Armed with a technically workable system
configuration
capacity
requirements,
and
which supports
other
projected
organizational
two items are prepared: an
RSSU"1P1ION REUIEW M,,~f;Pi ~I;t'l-I ~~::M"ITlri_ update to the original budget submission and

-
41221es
- Overdll r~u~1l Du'J,].t a documented update to the previous (i. c.,
<Sllmatet~USA ~17,,,~ las~ year's) version of the Systems
CAPACITY CONFIG.
PLflN PLAN 1 Env~ronment PL,l.n. The new plan is reviewed
- CPU -ENI/ PLAN by senior sys1:ems and corporate management
'50fS!I BUDGET
-PHIIPH -EIACI<I.!P
and used to support the budget requests
100Y - SOFT·... AAE -CHAtltlEL
-BI.!DGET REO ...... . us,." Budget
- OCEM" Buc~et presented at the annual Board of Directors

-
'51:22/"" <;17 '15 meetings in the Fall of each year.
S.£.P. PLRN
- formal Pia"
- Trac~Jng Mell'lo(lology
- Plan to ~eglons
6/1/S8
[ DUOGET REUI£WS The final steps in the process are to update
the Plan to· reflect any changes made during
the Board Reviews, and to reflect back to
~.£.P . REUIEW
1',1' 4/!l~
department managers, any impact on specific
requirements, which might have been
"
pon BORRD UPDATE
,/
identified during the initial research and
collection interviews.
- l.'p'l:It~ "f:~" ~ .. ~,,:

CONFIRMATION 1,° ~~s_cnsl 4.0 WORKLOAD CBARACTIl:RIZATION

,

S.E.P. Methodologv Primary source~~ of raw data for CPU workload

characterizaticn include the System

• 541 -

Find a CMG regional meeting near you at www.cmg.org/regions

 Learn the basics and latest aspects of IT Service Management at CMG's Annual Conference - www.cmg.org/conference

Measurement Facility /Resource Measurement VM!HPO DATA

IAVSlXADA1A
Facility (SMF/RHF) records for MVS/XA
workloads, and VM/ACCQUNT and VM/MONITOR
records for the VM/HPO workload. Additional
data is sampled from Omegamon, a vendor-
supplied real-time monitor which runs under
VM/XA-SP on the combined VM/MVS Development

Buy the Latest Conference Proceedings and Find Latest Computer Performance Management 'How To' for All Platforms at www.cmg.org
system complex. These Omegamon data are
analyzed periodically to determine the amount
of system resource consumed by the VM/XA-SP
system, which normally accounts for
approximately four (4) percent of the
Development system usage. DATA
COLLECTION
AND
REDUCTION
Join over 14,000 peers - subscribe to free CMG publication, MeasureIT(tm), at www.cmg.org/subscribe

" .1 Haaaurtupent Data Collection and

Raduot.ion

'ROFlLE
All data are summarized daily, weekly, and SAIl'LE
SELECTION
monthly except for the VM/XA-SF data which
are analyzed less frequently. Visa currently
produces workload measurement and tracking DATA
reports on a monthly basis, so the monthly ANALYSIS
sununary data (contained on Generation Data
Group tape datasets) are normally used.
Daily and weekly summary data are frequently
used to support ad hoc analyses and answer PC·SASED ANALYSIS AND PRESENTATION SYSTEMS
specific questions related to system (SPI'lEADSHEETSIGRAPHICS)
performance.
WORKLOAD DATA
COLLECTION AND ANALVSIS FLOW
Initial reduction of the MVS/XA data consists FIGURE 2
of running RMF summary reports to provide
data for the 24-hour system utilization
profiles described more fully below.
essentially the same for VM/HPO and MVS/XA
Reduction of the VM/HPO data is somewhat more systems, the first analysis of 24-hour
complex, as the host-based tools which are utilization profiles -- differs between the
used to reduce the VM/ACCOUNT and VM/MQNITOR two.
data execute under MVS/XA. Consequently, two
steps are required. First, the VM Monitor
Analysis Program (VMMAPl is run on a daily 4.2.1 MYS utilization Pro~il ••
basis, producing hard-copy, reduced data used
to generate 2~-hour VM/HPO system utilization
profiles. Th~ VMMAP results are also used to The MVS/XA profiles are generated by
support ongoing evaluation of system processing the RMF summary report output
performance. Second, the VM/MONITOR and records using a series of two VM EXECs. The
VM/ACCOUNT data are processed to provide a first summarizes the is-minute RMF intervals
monthly file suitable for input to the into hourly averages for the following data:
analysis tools which run on an MVS/XA host.
CPU utilization;

Demand paging rate and;

NUmbe~ of TSO users.

These steps and others inCluded in the data
analysis phase of workload management are The results of this program are then edited
shown pictorially in Figure 2, Workload Data to remove non-t~{pical days (e.g., weekend and
Collection and Analysis Flow. The data holidays for the Development System,
analysis phase begins just below the broken, Saturdays and Christmas day for the
horizontal line and consists of the Production System). Days during which
following components: significant outages or other aberrations
occurred are also removed. The second VM
EXEC averages the data remaining after
Generation of 24-hour utilization editing to crea~e a 24-hour profile of system
profiles;
utilization for the period analyzed. From
Generation of analysis data using a these profiles, average prime-shift and
host-based system accounting peak-hour values are extracted for later use
package; in converting workload data measured in CPu
hours to workload data expressed in percent
Creation of PC-based spreadsheets of CPU used or millions of instructions per
and presentation graphs. second (MIPS).

While the second and third components are

- 542 •

Find a CMG regional meeting near you at www.cmg.org/regions

 Learn the basics and latest aspects of IT Service Management at CMG's Annual Conference - www.cmg.org/conference

4.2.2 VM ptili1ation Profile' For the Development system, data are grouped
into thirty (30) departmental categories for
VM/CMS, MVS/Batch and MVS/TSO usage. The
VM/HPO system utilization profiles are primary measure of CPU workload data is total
generated manually using the daily VMMAP CPU hours consumed. Summary data grouping
reports. CPU utilization data on the departments along organizational division
detailed resource reports are inspected for lines are produced for monthly reporting

Buy the Latest Conference Proceedings and Find Latest Computer Performance Management 'How To' for All Platforms at www.cmg.org
profiles which are representative of system purposes. It is the detailed department
usage (e.g., no periods of exceptionally low data, however, which provides the foundation
usage during weekday prime-shift) and 5-10 of both for explaining aberrations in aggregated
the 20 or so working days for the month are Division data and for developing workload
averaged by hour using a PC-based spreadsheet forecasts. Detailed reports by individual
program. user ID and job name are also produced to
answer the inevitable questions which arise
when departmental workloads vary from one
Join over 14,000 peers - subscribe to free CMG publication, MeasureIT(tm), at www.cmg.org/subscribe

4.2.3. Account epd-. and Heming month to another.

ConytptiOA' for Workload
Charact_rization
Production workload usage statistics are
generated each month on the bas'is of the
Having generated 24-hour system utilization first 3-5 characters of the job name,
profiles, it is helpful before proceeding to depending on the level of detail required.
recognize the major differences between Twenty-seven (27) categories are currentl J
Development and Production workloads at Visa. tracked. These are sununarized into six (6)
The former are oriented toward user major workload groups for purposes of
department. The latter are related to presentation and projection.
specific applications systems which may be
supported by multiple departments. This
difference is significant because workload 4.3 ~••nt.tion of Workload Data
measurement data for the Development resource
are organized by user account code, while
Production data are organized by major system While much of the analysis of workload data
as identified by the job name. described abclve is accomplished through the
use of the detailed reports generated by RMF
and the accounting software, these reports
Consequently, the importance in such al. are not in a form suitable for presentation
environment of valid user account code and to decision makers. To make the transition
job naming standards cannot be over to a more suit.able form and to provide a
emphasized. These are monitored on an means to further manipulate and summarize
ongoing basis to assure, for example, that data for analysis, PC-based spreadsheet and
organizational changes and personnel shifts presentation graphics software is employed.
are correctly reflected in the appropriate The following kinds of presentation data are
system parameter tables. produced:

T:r~e effort required to develop and install Workload distribution,

b _h an account Jde structure and job naming
c:.nventions was :lOn-trivial. Monitoring for Workload usage trends, and
adherence to these standards is also time
consuming, but very important because the 24-hour utilization profiles.
correctness of such data has a basic effect
on the integrity of measurement data which is This information is produced for both the
the foundation of workload analysis at Visa. Development and Production workloads.
The department-related data, as shown below,
is used as a basis for projecting future
workload changes. If the data is not valid, 4.3.1 'l'be spread'M,t.
for example, including personnel from another
department, a manager asked for his
projections cannot respond since the basis Table I shows part of the spreadsheet used to
for his or her projection is inaccurate. The characterize the MVS/XA Production workloads.
same applies when improper job names are used The categories have evolved over time and
for Production systems. continue to change as new systems come online
and existing applications are withdrawn. A
real advantage in the use of spreadsheet
4.2.4 Ro.t-Bl,.d Inaly.is software is the ease with which the
of Sy.tem D.aa- Data categories of workload can be resummarized
Having created the 24-hour usage profiles and using data contained in the detailed, host-
verifying that data grouping mechanisms are generated reports. This can be done without
correct, the next step in analysis of significant reprogramming and without
workload measurement data consists of rerunning host-based job streams.
generating specific workload related reports
using a host-based system accounting package.
Such analysis is not dependent on any Table II is based on the same data as Table
particular vendor's software. At Visa, I, but is both further summarized and
however, a vendor was chosen which could converted from CPU hours to MIPS using the
supply a system capable of processing both average prime shift utilization calculated
VM/HPO and MVS/XA data. from the 24-hour profiles described above.

• 543 .

Find a CMG regional meeting near you at www.cmg.org/regions

 Learn the basics and latest aspects of IT Service Management at CMG's Annual Conference - www.cmg.org/conference

TAlILt I fMLE tv
""S.. tIIl PRODUC:rtOll ~OI'J<LOAD'S (CPU 1I0lJl'1$~ LAST I.lNlATo;.,
'S-31-~6 LAST UPD/lT£,

1·····_···..·..······_·····_··..·········_····_ ·..·.._·· -
• "<'!I-'IIA ~UCTlOt< •
_ . 1----..- - _...._--------_·_......· _...... _·__• .._ · _ · · ·......_
• V""'>IPO .. 0 ......0l'l0 <H~.' <x:Te1 ....._1 OCC" J<>l<H
5_:5·00
..••••••••••_ ••••
l'[Me'
1
I IlOflKLOlID (II~.)
· _• • • _ • • • • • • • • • • • _: - -• •
OCTe7
~._.
IKIV87 ~C01 .-00 "[1188 "RRBB
._____________ - - - - - - - - - - - - - - - - - - - - - - - - - - - - - •. ---
APIl89 :
I--------~·------------- -------------------------.~--~---~--------~:
1IIE:pRRtMNT R 70.9 ..:r.S $2.0 91.11> 63.2 :
tlll'PLICIUJGM ..,.OUP A ~1.1 37.1 ' .... 5 I GIIDlII' 111 H.'" "2.t> 21>.8 ".1 ~.7 :

....... ...., .....'.'..,

1 ""pLrt.;Atl~ A'
~ ~:~ ~ 2::~ 2~:~ ~t~ ~::~ ~
39.3 32.2 "7.9
,
1 APpLICIlTIOIoI A2 1.8 'I.... 1&.1& It:
• , ,.., ,
.. a.a :

Buy the Latest Conference Proceedings and Find Latest Computer Performance Management 'How To' for All Platforms at www.cmg.org
18AtCH IS7.1I 211 •• 1'57.0 I ll£PAIITHDlT II
•I '"
,.'
,., ,
...,.,... ....
L GA(>lII' III 1.0 :
Il~LICAflON ..,.OUP e '.' ~ ..... j 1.2 :

......
I GAD........

·
••
••
AppLJCllflllN GROllI' C
-.rCIlTION C.
-.rCArlOfO C2
... M.'.....
73.0
.. 7.2
22.7
~.,

~
'59.1
'11.'S
,.,
1".~
31.7
13."
1.....
""~
I~."
lb.'"
•
I DEHATIII!:Mt C
I .... OUp.;1
I G1l0UP \;2
, .. .. ,
,,
7.$ ,.'
......,,, 7.<f :
2 •• :
0.6- :

·•• "."."'
•
APl'\.lCFlrlOt< ell

..,,....,
'.' '.'
·
I GllOUP C.>

..........,..,, ,...'.'....,
'" .... <f :

... ...... ,........., ...... .. .....,

.......
,.......,
n.t

...
DTIKR llATCH 57.6 58." 4 ... 5 SO.S "2.7 1I"'.7 I llU'RflTIIEMr 0 12.7 16.$ 10.3 ..... 5 21.2 :
~. U.s ".
""
"'.O
12.0
,., ,
10.5 10 ••
'.'
I GlWUp 01
o. ,
..,,
M ~.3 '.1 ,
'" I GIIOUP 02 0.2 ,

·•• -...", ...,,.... ......'.'

h.2

..........
3.7

..,....., ..........,.
I GIlOUP 03 0.1 :

...'"'.' ............ '", ...,.. ..'.',.., ..,

2.9 3.E.
~:~
,~:~,
I GFlOUp 0 .. 0.3 :
~ M
.., '" '.-
M
I
I
GROUP De
GIOLOJP (0"
0.'
0.;
0.2
1.1
,
,
-.'
.
".2

...,.. .......,, ...,.....

M I GIKAIf' 07

·•••• "" .. .. ,.. · ......

".0
Join over 14,000 peers - subscribe to free CMG publication, MeasureIT(tm), at www.cmg.org/subscribe

I CltClUP De IC.)
'.' " 0." ".'<
"'~
12.$
~

.., '''''.1...
17.3 I ...... 11.3 1".(1
2." ,.,
13.1 Ih.:!
·'.EI
1 0TH[R OfPR/lTllENT5
t WlOUP 01
0.' ..'", 0.'
,.. ,., ... ,.. ....
1 IloAOUP 02

...... ,.. ....., ...... .....,

1511A~V5""H" 13.~ 10.1 19.~ 1•• 3 1".7 1•• 7 Ih.l 1 o1AD~ 03 Q.:l
\ CIC5 7.~ 1(1.3 11.1 ID.~
,.,
In.S 1 GFlOlJP"~

.""""
1 TSO 1 ..R(lup 0'"
• ..,
.........
I lOROUP 0 ..
I HIlIIT[O TIISKS, O€TllOllI(
• ". 20.0 1$.3 1'iI.3 t GRbUP ...7 '-'
\
\
\
STII.
5TM2
,TM3
W.
".
". ".
"" ""
""
""
""
""
""
" •
I D(PFlIITIlf:NT E
I GADUP [1 .........
12.$
.......,,
u ....
.........
1• . •
...
,..,..
17.&
..,
13.0

•
, STIIIIHO Tl'lSKS, OTHER
I 5HlI
\ ST02
17.~
1....
2.\
1"'.3
3.0
I.S
3~.3
~.~
3."
I~.'
2!i.~
3.'
2.3
17.5
~.'l
2.1S
17.9
'.5
•• S
2D.~
3.3
1.7
, GROUP f:2
1 GROIJl> [3
I GRGlU' f:"
I GROUP £5
...'.'... .....,.., ..,....,, ... ,.....
L'

.'.''.'.. ......,,.,..
\ 5T03 '15.11> 11>.5 9.9 7.2 ".1 5 ... 1 GROlli' f:'" 1.~
5TD~ ... ~ •• ~ ~.~ ~.'"
•
....., .....'.',
I 2.'" 4'.7 1.8
I "ISC
l _ ·__ •

L·_·······.....···_·····__··_·_··..··_..
I TOTRI.. ""S"'IIFl f'ADD 1Ill~:
• • : __ ··~·_~

189.3
,.~

2n.1
1&.7
~
10.S
33".0
7.5 2.2
--------~----~--------------

:3 .. 8
----------------..--
2~1.f>
2.6

238.5
~.2
.. ---
27'1.8
IlIIVHPO S"'$T SIIC HflCM
·• W,
W.
2 •• 1 27 ...
'-'
.....,
31.1
12.0

'.'
"11.0
13.0
:0'7.2

... ...
•1 w.
,.. .....'.'
5H. M
• W, I ... '.7
..,
,,..
b.o;
• W. ,
·,.__ _._-_ M ....~
tAIIL~ II • ""7
HVS/MIlI pROIIUCTlDM UllRI<LOI'IDS aups) I lIIVW'O FlpPue we HlleM
,.,
7.2 ~ 3." :
0/._._
2 "VS-'XI'I l/Ofl"'LOFlD
~~-------- ~----------~~~.:
21IPPLI(.ATlOH",
Z>'!p.... ICATl""....
.._._ _
~"'Ips..
...OCtd7
(I.'!I1
1«11>'87
_._-------_._
0.0'"
0.1'!1
_
.....D[Cd7
1.01
0.'2
_ _
Jl'lKEl8
__.._ __..
F[fetl
_-.~._---~-~--~----.--------------_.-.-~.-:

(1.1.
(1.01
0.'-"
1.(1. o.e..
I.~S :
0 • ..,
_00

<).13
_ .
IlPR88

Il.''!o :
·
I
~,
.. .. ~._ .... _.: -.._. __
T(lTAL IIn... Mpo ....' 1.0."
._~--~--.-.------------------------:

1.3.2 132.6 '~".1

~ 3 •• :

143.11>'

2""'~rtI'l1'lQN GIl\l<Jp 0 0.03 0.... 1.0' D."3 I.n 1.11 1.03 :

~"pPl.I(AnON (;Il(llJp e I.:n '.00 ~.DiI> I.~ 1.05 'J.7S D.'l3 :
2lJHI£Il OFltCH I."~ I.~.. 1.3" I."'; 1.~7 L3~ 1.7\
~$U8Sv~T~"~
~~TIIllT[O
25TIlIlT[O
t ......-.:S,
rR~S,
"HUOR~
DU'i[R :
o.~~
0.::>3
0 .•3
o.~~
0.21
0 •• 0
D....
0."0

(I.n
(1.3"
0."""
D.W
.'.35
0 ••7
0."2
').35
".5;;'
o .• ~
(1 •• 5
0."
0 .• 7
:
:
:
.--------~.----~----------: --~-_._~~ .. --... ------------------------------~-----~-~--------: month during prlme shift. This data is also
~ ", ..", "5l!~ 0;.7 ".n 7.1. 5.0;., 5.03 !;.H ';' ••0 :
summarized and converted to MIPS in a fashion
l'Ilp3 UHUSl:O
•2 -- •• ~ _._.~ •• : ~
(13 .... 9..
~~
3.5.. ._. ...>7 •.• <1,
__ ~
<;.(>5
•• ".l(l :

•
... n
TMRL l'UpS : 10.70 _ . _ . 10.70 10.70
.. __ _
1(I.ro 10.70
._
\".7" '0.10:. similar to that described above for
Production workloads.

This is done by simply apportioning the 4.3.2 The Graphics

measured utilization on the basis of the CPU
hours consumed by each workload shown in
Table I. Using summarized data from spreadsheets such
as those described above, graphics are
produced each month (or when requested) for
The information shown in Table III is used to both Production and Development workloads.
track other parameters of interest, such as Figure 3 shows system utilization profiles
number of TSO sessions, number of batch jobs for the Development VM/HPO and MVS/XA systems
and peak-to-prime utilization ratios. for the month of February 1988. Although
MVS!XA data were generated in Greenwich Mean
Table IV shows the departmental workload
split for the Development VM/HPO system
measured in total CPU hours consumed for the
DEVELOPMENT SYSTEMS USAGE PROFILES
FEBRUARY 1988

...., F III 200 ~3090-':"'::":'200""'Uc:T1"'LIZA:o::::Tc:ION::::.:('-PE::R.::C"'E:::NTl"----- _ _- _ - - _ - ,

HV~"'K4 PIIOouctHIII

,
I
_ _._
IIVS HR PA001,l,'tIOH
~-_
Sl.Jpl'I.f:H[NTFIII~ _10:1-000 TI•

.. ~_ ~._
...-:Io:r HG ClAfA
_.-._
LMf IJP1lllr[,
5-l1_9$
----~ _ .
150
I
1 _PLEI\[H1'AII;~ ClAtA ··_: _. Oct07
__ • • • • • 110"'87 ot:c,,7
•••••• •••• ~ .~
.-0 rUllle
__ _____
~~
0$ ~ ~ ~_~
..,1100

II'WS.I1<A tOfRl..
I ~(II1TM (Hour.) •. n 1 ••1I D."" 1.1'"

·" ... ... ...

1.7'!1 1.33 1.";

... '.M..,
I CUIIUL GIIDlolTN , ......... ' 1.23
100
I
I
1101,111;$
$[5SIOM5 .~ ,,~
..,
,,~ 125(' 1~~
1 """"TH <-......> '.Of, D."'iI 1.,;'tt

·
1

1 8..TeH
1 HOURS
1
'l"O"T~

.J»"s
,:;.,..u ..... ' ~.rl 1.0' 1.0' 1.16

1".7
,,")'!''>
111>.1
e~".
50

1
I
•,r"'J
<"'O"T>I <;_~.)
'>RDIIT>I , ....... J
1.00
1.12
1.09
0.-;1" .1-....._1Iii
I
I
I
DAH.
noDEL
£"tps
'''JIIE SHrFT A"'~ UTIL
".
10.?
.11>.1
.W
10.7
S$ ••
".
10.7
$ •• li'
0;00 3:00 6:00 9:00
AVERAGE BY WORKDAY HOUR (P.S.T.l
12:00 15;00 18:00 21:00

I P[Rt~ HOUlI' AVO ...lIL ee.o "'S.7 7•• 1

L.'"
I '£"10:,1'101"'1: AFlTlO '.SS 1.3"
I
·•
I l:RIp5... IIOHTHLV CPU IIA

I PRIHT ""TR
t PRIII[ SHII'T LI1'IIES HIli 11...11
2.11

HIll lilA
2.35 2 •• '

H,"f< tVlI 1 11
[ iBitI VM/HPO

FIGURE 3
0 MVSIXA

1-···············..··.._·..······..··..··..···_······_······
L plUIIE :>111'" pFlGE:> : MIll MIFl H/A tVol II..... IVA
_ "
.

• 544 •

Find a CMG regional meeting near you at www.cmg.org/regions

 Learn the basics and latest aspects of IT Service Management at CMG's Annual Conference - www.cmg.org/conference

Time (GMT), used to synchronize systems

worldwide, times on the X-axis were converted DEVELOPMENT SYSTEM WORKLOAD
from GMT to Pacific Standard Time (PST) to OFF SHIFT MVSIXA USAGE
coincide with management's PST orientation FEBRUARY 1988
toward the Development system. The Y-axis is
shown at 200 percent, representing 100

Buy the Latest Conference Proceedings and Find Latest Computer Performance Management 'How To' for All Platforms at www.cmg.org
percent for each of the two processors in the
3090-200 complex.

Production systems, on the other hand are

perceived more on a GMT basis, so data is
presented using GMT as the X-axis measure.
Use of a PC· and presentation graphics makes
such conversions very easy_
Join over 14,000 peers - subscribe to free CMG publication, MeasureIT(tm), at www.cmg.org/subscribe

The pie charts shown in Figures 4 through 6

indicate the percentage of total usage
represented by the various Development user
departments. Figures 4 and 5 show prime
shift VM/HPO and MVSJXA usage respectively I ,,,
DEPARTI,£NT E

while Figure 6 shows the MVS/XA off-shift

workloads.
FIGURE 6
DEVELOPMENT SYSTEM WORKLOAD
PRIME SHIFT VM/HPO USAGE
FEBRUARY 1988
Figure 7 shows Production system workloads at
the San Mateo facility. Notice that these
workloads are quite different from
Development and are based on the job name
data described above. Proportional pies
indicate the relative amount of resource
OEPARTUENT D consumed during the prime and off shifts.
=
The last workload chart, shown in Figure 8,
indicates the trend in Development VM/HPO CPU

DEPARTMENT A

"" '"
-
DEPARTMENT C

DEPAFlTlo£NT B

PRODUCTION IMVS/XA) WORKLOADS

JANUARY 1987

FIGURE 4

DEVELOPMENT SYSTEM WORKLOAD

~.-'
PRIME SHIFT MVSIXA USAGE APPlICA"lllJltB
OItI~~TDl
FEBRUARY 1988 ....

"'PPI.~IQN"

OFF SHIFT
(1800-2«10 GMT}
-
IA.8nSTOCIISTAIlTl'D

DEPARTJENT B PRIME SHIFT

,,% (0lXl0-1800GMl)

OlllER DEPTS
,%

DEPARTMENT E
,,% FIGURE 7

DEPARTMENT D

'" resource consu~ed by the various departments

during the pedk hour for each month shown.
In this graph the CPU hours have been
converted to percent CPU utilization. It is
used to watch for shifts in peak-hour usage,
FIGURES when the system is most heavily stressed.

545 -

Find a CMG regional meeting near you at www.cmg.org/regions

 Learn the basics and latest aspects of IT Service Management at CMG's Annual Conference - www.cmg.org/conference

processor resources. Preparation is

1988 DEVELOPMENT WORKLOAD extensive in Jrder to provide the data
VMlHPO PEAK WOUR USAGE TRENDS
required and make most effective use of
,P~E~R~CE::NT:":U,,TI,,:L1,,:ZA:::T~ION:::.!(c:112~3090-200~~I~_~ ----, management t irr,e . The interviews are
invaluable for at least the following
"" " reasons:
80

Buy the Latest Conference Proceedings and Find Latest Computer Performance Management 'How To' for All Platforms at www.cmg.org
60
A personal contact and rapport is
established which is not possible
using a written survey;
2<1
Questions and concerns can arise
o and be dealt with quickly and
OCT NOV DEC JAN FEB MAR APR MAY JUN JUl AUG SEP OCT NOV DEC
effect i vely;
Join over 14,000 peers - subscribe to free CMG publication, MeasureIT(tm), at www.cmg.org/subscribe

I 87 I 88 I

am oePTB ~ OEPTC Staff pere0nnel who may be the

JllIII OTHER DEPTS o OEPTE actual se .;e of information, are
often ~: ltroduced by their
management into the process,
FIGURES providing valuable sources of
understanding of the systems and
organi zation;
Suggested improvement s to tb<.:
5.0 WORKLOAD FORBCASTJRG process are frequently offered; and
Non-technical and/or organizational
The forecasts of workload growth (or issues can be discussed; these may
decrease) are based on interviews with key be critical.
department managers and the use of historical
trend data. However 1 the specific methods After completing each interview, a detailed,
used to forecast Development System written follow-up is sent to each department
departmental workloads are somewhat different manager, indicating the results of the
from those used for Production workloads. interviews as understood by the capacity
Consequently, these are addressed separately. planning personnel. The written follow-up
provides a clear understanding and agreement
of requirements and is also used as a vehicle
5.1 WorklQad Forlcalt Inttryi•• Proce•• to obtain concurrence by the next two higher
levels of management. Everyone has an
opportunity to be involved early on,
Each year extensive personal interviews are providing a solid foundation for future
conducted with approximately 30 department projections and configuration planning.
managers. These managers represent systems Misunderstandings are minimized. It works!
development, systems engineering, product
development, finance/accounting and corporate
user functions throughout Visa. The 5.2 D.ytlopmant Workload rorleaat
interviews are structured to last
approximately one hour and cover the
following topics: The graph shown in Figure 9, along with a
similar graph covering the previous 15 months
(not shown) provides the basis for estimating
Review of past 15 months resource future workloads. The IImessy" graph in
usage; Figure 9 starts out "clean" and is marked-Up
by the interview participants. The line
Resource requirements projected
for the next 12-24 months,
depending on the resource; DEVELOPMENT SYSTEM USAGE
Vt.lJHPOWORKLOAD FORECAST WOAKSHEET

Estimated delivery of projected new Go

60
applications;

.
~(! <;f~
50 ~

architectural changes cp
Software ,.7'.
which could affect existing U

systems; H 30
0
U 20
R
volume projections for Production S
systems; "o:IWWrt'-~,+,I..;J,.,..4..,-h-....J..,4-.-rm-rrtl,,.;W
OCT JANAAA JULOCT JANAPRJUL OCT JAN APR JUL OCT JAN APRJUL
Other items of interest to the 187 I 88 I 89 I 90 I 91
department manager.
I . DEPART A1 (Yl/TANG)

The interviews are conducted by capacity ! -- Ac:Wal Usage 0 Planned Usage

planning personnel accompanied by a FIGURE 9

representative from Systems Operations. They
cover all major subsystems including CPU,
OASD, tape, workstation and co~unicationa

Find a CMG regional meeting near you at www.cmg.org/regions

 Learn the basics and latest aspects of IT Service Management at CMG's Annual Conference - www.cmg.org/conference

shows several recent months of actual usage

for a particular department, while the bars PRODUCTION (MVSIXA) SYSTEM USAGE
indicate the results of last year's 1988 PEAK SEASON PROJECTION
projection. Experience has shown that while
managers have a difficult time projecting
absolute values of resources required (there
is currently no charge-back system and

Buy the Latest Conference Proceedings and Find Latest Computer Performance Management 'How To' for All Platforms at www.cmg.org
hardware/software resources are budgeted
centrally) they are very good at projecting
requirements relative to what was previously
used to produce a known level of work
product. PiGLures also help a great deal.

Having made pro jections in such a manner it

Join over 14,000 peers - subscribe to free CMG publication, MeasureIT(tm), at www.cmg.org/subscribe

is relatively easy to "clean up" the results

for later presentation (via the follow-up -~
~

memo) and for data-entry into a spreadsheet PEAK SEASON 1987 PEAK SEASON 1988
used to aggregate all projections by (3090-150) (3090-150)
organizational entity. The result of a
typical cleaned-up profile is shown in Figure
1 o. The seasonal peaks depicted are very

FIGURE 11

30 percent higher than an average day.

Hence, trending workload growth on a monthly
DEVELOPMENT SYSTEM USAGE basis would not be particularly useful for
VUlHPO WORKLOAD FORECAST
projecting annual capacity requirements.

~
=1
S<l
Growth rates for individual applications
systems are applied to the base (i.e., last
year's peak), result.Llg L: the ri ght-most
u
H
" I1orlnlon"lll mnnnrlhrrnnniII Illrlnnnnr graph on Figure 11. The use of a small
o "
U spreadsheet, ,s:hown in Table V, makes it very
R "
S easy to try different growth assumptions and
10 then determine the effect on projected
,-'iJ\Il.f\I\l\!fI\I\I4_AA
OCT JAN APR JUl OCT JAN APR JUl OCT JAN APR JUl OCT JAN APR Jl.Jl
1871 88 I 89 I 90 I 91

r.. IlLE v

FIGURE 10

difficult to obtain without the use of

graphics due to the use in verbal
descriptions of terms like "more" and
"bigger" .

5.3 Production Workload roreca1t requirements. The growth rates are based on
two factors:
While Production workloads are tracked
monthly using the graphs described in section Information obtained during the
-4.0 above, the projections are based solely interview process; and
on data obtained during the seasonal peak
periods. This contrasts sharply with the Esti.mated growth in business
Development workloads which tend to vary more volumes obtained from product and
continuously over time. The basis for corporate planners.
Production workload projections is a
composite workload profile developed from Projection models exist for the most
measurements taken during the three-day peak significant applications. These models
period of the primary ;=.pplication, shown as convert forecasts of businebs element growth
Application A on Figure 11. The volume of (e.g., number of authorization transactions
data processed by this system on the four or or number of_ sales drafts) to an equivalent
five peak days of the year can be as much as system usage estimate. In addition, a new

- 547 -

Find a CMG regional meeting near you at www.cmg.org/regions

 Learn the basics and latest aspects of IT Service Management at CMG's Annual Conference - www.cmg.org/conference

workload may be estimated and inserted into Workload Service Levels. The goal is to
the projection if a new system is scheduled measure system workloads relative to
for installation. Results shown in Figure 11 agreed service levels. While system
indicate the need to either off-load utilization is a reasonable gross
significant workload or upgrade the 3090-150 indicator, a high aggregate system
to meet peak-season requirements. An upgrade utilization is undesirable only if

Buy the Latest Conference Proceedings and Find Latest Computer Performance Management 'How To' for All Platforms at www.cmg.org
was recommended after a careful analysis of service levels are in jeopardy.
each workload shown.
Forecast; ng Methodology. While the
6.0 WORKLOAD TRACKIlIG interviews and managerial experience
will never be replaced by statistical
analysis, some form of statistically
The workload tracking process at Visa can be based trending is desirable to further
Join over 14,000 peers - subscribe to free CMG publication, MeasureIT(tm), at www.cmg.org/subscribe

described in two categories: monthly usage support the goal of obtaining the best
reporting and biannual (annual for possible projections. Such seasonal
Production) reporting of projections vs. trend projections could be used as a
actual data. The first category is comprised pc:nt of departure in the interview
of the usage charts shown in Section 4.0 process.
above, along with others depicting TSO- only
and batch-only for the Development workloads.
In addition to the graphs, summary reports by System Modelling. Where possible, based
department and detailed reports by job name on increased understanding of workloads,
(Production) and user 10 (Development) are the Installation Control Specification
provided for use by managers who wish to (ICS) and Installation performance
understand their department's application or Specification (IPS) parameters in MVS/XA
usage characteristics in more depth. Such SYS1.PARMLIB will be modified to place
reporting allows managers to monitor the each worklcad in its own performance
workloads at their own discretion. group. This will support the modelling
of at least some of the MVS/XA workloads
using one of the available vendor-
The second type of tracking is done bi- suoolied oackaqes, and will allow a
annually for Development VM/HPO and MVS/XA rough prediction ot performance under
departmental workloads, and annually (since differing workload assumptions.
projections are made from annual-peak to
annual-peak) for Production MVS/XA workloads.
These areas for future effort are currently
being assessed for relative priority and will
For Development workloads these actual VB. be scheduled to begin accordingly.
plan reports consist of graphs similar to
Figure 10 but with the Current monthly actual
data shown. In addition, high level
aggregations by organizational entity are
developed for senior systems management. 8 •0 SlJMMIIRY AlII) COlICLOSIOlIS

For Production workloads a series graphs such Previous section~; have intrOduced briefly the
as shown in Figure 11 are presented, but the systems hardware and software environment and
left-most pie shows actual data for the same the systems planning process currently
year. Since this is the first year of making evolving at Visa Intern~tional. The process,
projections in this fashion, it is not yet managed centrally, appl~es to all three data
possible to show actual results. centers worldwidl=:!. Elements of the process
can be sumroarizeq to include:

Data collection through reduction

of syatem measurement data and
7 •0 IVT1lI\IC PLANS extensi.ve personal interviews:

Formal, written follow-up and

Although a significant amount of effort has reviewE'. of planning assumptions
been expended to develop a process for with management decision makers;
actively managing the system workloads in
Visa's three data processing centers, a great Direct links between the planning
deal remains to be done. The following areas and bucgeting processes;
will be addressed during future planning
cycles: Preparation of workload
churacterization data using host-
based SOftware;
Measurement Dat.a Base. A database of
performance data is being devaloped Summary and presentation of data to
using SAS and vendor-supplied management using PC-based
application software. This will provide spreadsheets and presentation
a more efficient access to consistent graphics;
system measurement data. Current
reporting mechanisms will be extended Support for system configuration
and/or replaced. planning;

• 548 -

Find a CMG regional meeting near you at www.cmg.org/regions

 Learn the basics and latest aspects of IT Service Management at CMG's Annual Conference - www.cmg.org/conference

Documentation of results in a Budget Input. At Visa the budget detail

formal Systems Environment Plan data is assembled using PC-based
which is modified annually to spreadsheet "shells" which are completed
reflect changes; by department managers. More detailed
supporting schedules are maintained to
support high dollar amount line-items.

Buy the Latest Conference Proceedings and Find Latest Computer Performance Management 'How To' for All Platforms at www.cmg.org
Feedback mechanisms and ongoing
reporting to update department One of these line-items is data
managers. processing Hardware Expense. For the
1988 budget cycle, the initial version
this detailed support schedule was
The use of PC-based graphics is an integral actually produced for the budgeting
and key part of this process. Workload staff, by systems planning personnel at
management characterization forecasting their own initiation. The response was
and modification -- is directly influenced by very positive and less time was spent in
Join over 14,000 peers - subscribe to free CMG publication, MeasureIT(tm), at www.cmg.org/subscribe

the use of these tools. communicating technical requirements to

the bUdgeting staff.

Throughout the evolution of workload

management elements of the planning process, Applications Capacity Plans. In a few
several conclusigns have been drawn. These instances, applications development
include: departments perform capacity planning
efforts for selected applications. Such
efforts mayor may not be formally
value of Persgnal Interyiews. The value documented. Nevertheless, the results
of personal interviews with department of these efforts are incorporated,
managers cannot be over- emphasized. either directly or by reference, into
Not only is a rapport established which consolidated capacity projections
allows easy follow-up, but the arising from the processes described
conversations inevitably uncover above. Since these consolidated
requirements which would otherwise be projections are a part of the overall
missed. In addition, capacity planners Sy~tems Environment Plan, the end
conducting the informal sessions learn a results become more firmly supported
great deal about the environment, from both a technical and organizational
applications, and personnel which viewpoint.
comprise the organization.
The workload management effort described
Formal Follow-up. While time consuming above has been evolutionary. The process
to prepare, t'he memoranda used to verify continues to change as planners become
and document stated requirements are skilled at using new tools and techniques.
invaluable. Distribution to senior While these tools and techniques can make the
managers regarding the needs of their process easier, they are not a substitute for
staff management team causes the team the personal interactions among planners and
members to take part ownership in the line managers.
process. This causes the estimates and
projections to be more thoughtfully
developed. Beyond these positive
aspects, the follow-up provides a
common, documented source of projected
requirements which can be used to track
against actual requirements.

Communication. The interactive nature

of developing workload profiles is an
excellent tool for communicating with
management. In some instances,
perceptions about when peaks occur and
which workloads cause them, are based on
historical experience. As the
environment changes, such perceptions
can become outdated. Providing an
updated view of recent history results
in better decisions made by more
informed decision makers.

Quick Response. The use of PC-based

tools has made it possible to respond
very quickly as requirements change and
needs for communication arise.
Presentation materials can be developed,
usually, without the need for time-
consuming reprogramming efforts
sometimes associated with host-based
systems.

- 549 •

Find a CMG regional meeting near you at www.cmg.org/regions