0% found this document useful (0 votes)
8 views12 pages

Sizing Applications

The Computer Measurement Group (CMG) is a non-profit organization focused on the measurement and management of computer systems, emphasizing performance evaluation and capacity management. This document discusses the importance of Application Sizing (A.S.) in the Systems Development Life Cycle and the challenges faced in coordinating A.S. across departments. It highlights the need for collaboration among various stakeholders to ensure accurate resource planning and performance management for new applications.

Uploaded by

Sumukh Mullangi
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
8 views12 pages

Sizing Applications

The Computer Measurement Group (CMG) is a non-profit organization focused on the measurement and management of computer systems, emphasizing performance evaluation and capacity management. This document discusses the importance of Application Sizing (A.S.) in the Systems Development Life Cycle and the challenges faced in coordinating A.S. across departments. It highlights the need for collaboration among various stakeholders to ensure accurate resource planning and performance management for new applications.

Uploaded by

Sumukh Mullangi
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 12

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 1991 International Conference.

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

Copyright Notice and License

Copyright 1991 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

SIZING ApPLICATIONS BY COOPERATING:


MASTERING THE UNMEASURABLE

Buy the Latest Conference Proceedings and Find Latest Computer Performance Management 'How To' for All Platforms at www.cmg.org
Dr. Leonard D. Lipner
BGS Systems, Inc.
Abstract
Three observations clearly indicate the importance of Application
Join over 14,000 peers - subscribe to free CMG publication, MeasureIT(tm), at www.cmg.org/subscribe

Sizing (A.S.) throughout the Systems Development Life Cycle, from


high level design onward:
I. No matter how fast the power of computers increases, new
appl ications' demand for resources increases even faster.
2. A large part of new development is for highly visible
"mission-critical" systems.
3. Productivity enablers, plus increasing development skills within
user departments, dramatically shorten the development life cycle.
In many shops these realities lead to an ironic contradiction:
everyone agrees that A.S. during design is important, but also that
it's someone else's job.
The contradiction - a variation on the old theme of non-communication
among departments - isn't new. It's jlJst making itself felt more
strongly today than ever before.
We explore the problem to expose its sources, and to distinguish
between real barriers and simple excuse-making. From what we find, we
offer a few ideas about moving beyond barriers to successful
performance, cost and resource planning for new applications.

INTRODUCTION: ONCE UPON A TIME ... control, so he's your biggest supporter.
Let's say you're the Capacity Planner Your first assignment is to size a
for your organization, and you've just couple of highly visible and
been given the related job of performance-critical new applications,
coordinatin Application Sizing (A.S.) both of which are just entering the
during the aevelopment process. It's a design stage. The new responsibility is
function that's never been addressed supposed to take about 1/3 of your time.
before at your company. The Director of
IS Operations understands the benefits On the other hand, it's clear to you
of A.S. (in principle), mostly because that coordinating A.S. is really
he's been burned by its absence before. different from the "individual
Specifically, when a couple of recent contributions" you're used to.
new applications went into full Specifically, A.S. entails cooperation
production, they turned out to be much by people from several other
bigger than expected, and a seriouS- departments, and in your organization,
capacity shortfall resulted. Given that's a problem. The backing of the IS
Operations' responsibility to meet Operations Director helps, of course.
production service levels, the director But it doesn't imply that anybody else's
wound up with "egg on his face". A.S. time is committed, and he's not really
promises to bring all that under able to establish that commitment.

977

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

To understand and size the two chosen IlIU1M11' IS I'lllSOHIII: 111I [11\' 01
applications, you'll have to gather CIIIIRIIC II RS
information from designers and
developers, and from some others. What's Any way you luuk al it, your job j'. 1.'1
more, since successful A.S. requires provide information tu help mana~~r~ an~

Buy the Latest Conference Proceedings and Find Latest Computer Performance Management 'How To' for All Platforms at www.cmg.org
periodic refinement during each phase of executives make a variety uf important
the development life cycle, you'll have decisions: go/no-go decisiuns abOIJt ttl~
to seek those people's help more than application itself, implementation
once. decisions, capacity and budget
decisions. You're in the business of
By and large, the people who have the providing not just any information, but
Join over 14,000 peers - subscribe to free CMG publication, MeasureIT(tm), at www.cmg.org/subscribe

information you need don't know what SOOd information, as accurate as it can
A.S. is, and they surely don't view it e. The best way to see who's involved
as part of their job. (In fact, so far in the A.S. effort is to ask:
it really isn't, but that's partly their
own choice~we'll see below). If they Whose day-to-day decisions
know anything about A.S. at all, they influence the accuracy of the
may still view it as an uncertain information you're supposed to
technology of dubious value, so you may provide?
have to deal with skeptics along the
way. On top of this, their schedules are To avoid the serious risks of being
probably over-full with too many wrong, it's extremely important that you
deliverables, promised too soon. Your take into consideration each and every
request for help may well be seen as an i nfl uencer.
unwelcome intrusion.
First, let's look at the information
Finding yourself in such an optimistic itself. Normally, the information you
scenario, what do you do? periodically have to report includes at
least the following:
Take heart! There's more than enough
advice, and a sufficient set of 1. The new application's resource
techniques, to introduce A.S. painlessly requirements - CPU time, memory, I/O
into your organization. It's a matter of activity, DASD and tape storage, and
training, for you and others. network traffic. The long-term view
(per day, week, month) is required,
The fact is, in all but a few as well as the short-term view (usage
organizations coordinating A.S. requires during peak hours). Some indication
interpersonal skills as much as is also needed of how these numbers
technical ones. It requires that you will change over time.
overcome cross-departmental barriers -
which may be a bit of a surprise, but The most reliable way to estimate
it's true. And if your technical skills these figures is to analyze the
exceed your ability to convince people resource consumption patterns of the
to spend their valuable time working major transactions and jobs in the
with you, you may find advice on how to application, and to extend that
do so helpful. information by the anticipated
workload volumes.
That's what this paper is about:
crossing barriers, enlisting help, and 2. Periodic cost - based on assumptions
(just maybe) creating a few converts. about charg~ack rates, and the
answers to 1. above.
By the way, there is ample literature on
the precise data required for A.S., and 3. The application's performance during
on how to analyze and report it once critic~riods - response times,
you've collected it (see, for example. ihroughputs;-ana to the extent you
[I] and [2]). This is a management- and can identify them, major risks to the
procedures-oriented paper, and will not installation's ability to meet its
deal with specific technical issues. service commitments. Deriving this

978

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

information usually requires a groups are the major players in the


performance model, along with some development project. Among the important
assumptions about the system and tasks assigned to them, the A.S.
network on which the new application coordinator needs to be especially aware
is most likely to run. of a few:

Buy the Latest Conference Proceedings and Find Latest Computer Performance Management 'How To' for All Platforms at www.cmg.org
4. The environmental impact in Interface with the end user, to
particular, how the application will determine the flow of work in his
affect response times and throughputs department, the role the
for current production workloads, application will play, the specific
with their own service level functions the application will
Join over 14,000 peers - subscribe to free CMG publication, MeasureIT(tm), at www.cmg.org/subscribe

objectives to be met. provide, and the workload volumes.


Most applications you're likely to Preparation of the application
analyze are inherently complex. Besides functional specification.
the obvious complexity of the hardware
and operating systems, they use (and are Design of logical data
influenced by) subsystems like CICS, requirements.
IMS, or others, and DBMS's such as DB2.
There is an enormous variety of Translation of functional
facilities, and an even larger variety specifications and logical data
of ways to use them. Different requirements into program
facilities require vastly different specifications.
amounts of CPU, I/O, DASD storage, and
so on. Most importantly, the major Anyone familiar with these tasks can
determinant of the application's easily appreciate the large number of
i"esou_rce requirements, performance and individual decisions these people make
costs is the choice of facilities used, which influence the application's
how they're used, and how often they~re resource requir'ements, and by extension,
used. its cost and performance. As just two
examples, analyst/designers determine
Unfortunately, there's no single person (I) the mapping of user-specified
or document you can turn to in order to functions into computer transactions and
build a good understanding of the jobs, and (2) at a more detailed level,
application's use of system and the information to be displayed on each
subsystem facilities. They are transaction's screen(s), and how many
determined by the individual choices of functions will be carried out when
a large number of people. screens are entered.
And of course, when you look at * Lead application coders
performance it's important to remember
that new applications usually have tu In general, no matter how tightly
contend with other work on the system, program specifications are written,
and in the network. coders have a qreat deal of latitude
when translating them into code.
Each of the people or groups in the Everyone who has ever programmed knows
following list may, in the normal course that there are normally a large number
of their work, influence the outcome of of ways to get a task done (internal vs.
the application development project. To externa 1 sort i og, sequent i a1 search i ng
make sure you haven't left out any vs. more clever methods, SQL variations,
important influencers, scan the list to etc.) Much of the variability from one
determine which of them you need to coder to another is due strictly to the
interview. coder's level of training, or
cleverness, or the development
* Lead application designers and data department's culture, traditions or
designers (more formal) programmi ng standards. It
can easily happen that important
It hardly needs saying that these two built-in subsystem facilities are

979

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

rewritten from scratch (for better or is responsible for installing the


worse) simply because the coder is distributed operating system and
unaware of them. subsystems, and selecting options which
tailor those systems to the
In most installations, it is ultimately organization's needs.

Buy the Latest Conference Proceedings and Find Latest Computer Performance Management 'How To' for All Platforms at www.cmg.org
the coders who determine which
facilities are invoked, in what ways and Installation and customization are in
how often, within a job or transaction. general done for the company as a whole,
Usually, there are far too many coders, rather than for a single application,
and even more details, for it to be and they're not done very often. As a
reasonab1e to collect them all . However, result, you won't have to visit them
Join over 14,000 peers - subscribe to free CMG publication, MeasureIT(tm), at www.cmg.org/subscribe

much can be gained from a brief very often, but it's important not to
interview with a lead coder, or senior overlook them. MVS option selections
programmer, in order to gain an (performance groups, priorities, storage
understanding of the general way coders isolation, page/swap threshholds)
go about implementing functions - obviously have possibly serious A.S. and
particularly the facilities they tend to performance impacts. Analogous comments
use repeatedly, and the point at which apply to installing and customizing
they develop functions from scratch. CICS, IMS, DB2 and other subsystems.
This sort of information may be useful Typical subsystem issues include
when you have to estimate, at a high- standard region and buffer pool sizes
level, the use of subsystem facilities (and ultimately hit ratios, with their
by an application. implications for logical/physical I/O
ratios), pre-loading of code and
* Data administrators / Data base crit i ca1 databases, and use of ESA
administrators facilities like VLF and LLA. Once again,
the results for A.S. can be large.
These people playa role for data
analogous to coders: part of their job * Computer operations management
is to translate logical views of data
into physical data files and databases. Here the issue is not so much the
Once again, their decisions can absolute importance of the application
seriously affect resource consumption. in question, but its importance relative
Individual data items mayor may not be to other production and test workloads.
grouped in segments or rows, and clearly As operating systems and subsystems
such decisions can have a strong mature, more and more facil ities are
influence on the CPU and I/O activity placed in Operations' hands to control
needed to read or write a logical view. the flow of resources to one or another
workload. Through these mechanisms they
Data (base) administrators also deal can decide on the maximum throughput of
with issues of data security, backup, each workload at times when resources
reorganization and recapture of unused are scarce.
space, embedded free space, and the type
of storage technology used for files and It's of course tempting to expect,
databases. Clearly all of these choices perhaps even demand, that critical
can materially influence application workloads be given all the resources
performance and resource consumption. they require, on demand. Often, an
attempt is made to do so. However
* Systems programmers and subsystems practical experience shows that,
(CICS, IMS, DB2) programmers particularly in tight economic times,
tradeoffs inevitably have to be made.
At this point, we begin to move away It's worthwhile for the A.S. specialist
from those directly involved in to be aware of any possible limitations,
application development, to groups whose and to account for them when predicting
influence is less direct, but application performance.
potentially just as dramatic. This group

980

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

* Capacity planners less accurate about resources, costs and


performance.
This is perhaps the most obvious group
of all, since they recommend, and
therefore strongly influence, the kind A PLETHORA OF OBJECTIONS: THE PLOT

Buy the Latest Conference Proceedings and Find Latest Computer Performance Management 'How To' for All Platforms at www.cmg.org
and amount of resources to be physically THICKENS
available to application systems
overall. (Capacity Planning and It's now time to set out on your A.S.
Application SiZing may well be done by journey. One way or another, you've
the same group, but if not, there's gotten hold of a brief desciption of the
clearly a need for A.S. to seek help applications you're supposed to analyze,
Join over 14,000 peers - subscribe to free CMG publication, MeasureIT(tm), at www.cmg.org/subscribe

from the Capacity group). so at least you've got a starting point.


As you survey the landscape, you see the
* Computer systems financial management Director of Operations, smiling at you
encouragingly. But you also see a
The main information sought is an beehive of activity - all those
estimate of future chargeback rates, influencers busy at their jobs,
used to estimate the cost of the oblivious of what you're up to, and
application. Related to this is the cost unaware that you're planning to ask them
of new hardware (and any deals, such as for anything.
volume discounts, that may apply), if
you want to determine the feasibility of All of those people have their own
an alternative architecture to provide biases, their own jargon, and their own
the functions the end user says he notions of what's valuable. So first
needs. recognize that you're going to have to
tailor your approach to each of them.
* The end user
To be sure, some of them (the naturally
In most cases, the application analyst curious ones) will be interested in what
deals directly with the end user, and you're doing, and may even support it.
represents him in discussions with A.S. They'll be especially interested and/or
However, A.S. sometimes uncovers helpful if you can tell them clearly
high-cost functions with marginal value what you're looking for, and if they can
to the end user, leading the user to provide it easily. For instance, that's
consider whether they're really worth probably the reaction you'll get from
developing. Such negotiations sometimes systems and subsystems programmers -
take place directly between A.S. and the because there's really no guesswork
end user. there. You're looking for some
information about the way they set up
These are the main players - though the system, and that's just not
there may occasionally be others. When controversial. Besides, by and large,
all the information and people are they're intrigued by strange new
considered, it becomes clear that at the technologies.
outset of development it's almost
impossible for anyone of them to have a But you're going to run into resistance,
complete view of the application. So perhaps more than you expect. In the
sizing the application requires working author's experience, there seem to be at
with all, or at least most, of them. least four motives making people
reluctant to sign on fully to the A.S.
Articulating and communicating a total effort:
view of the application, for review by
the organization as a Whole, is one of I. Misunderstanding of the purpose and
the most obvious, most immediate, and methods of A.S., especially
most valuable benefits of A.S .. performance modeling.
Interestingly, it isn't even an explicit
goal of A.S. - simply a "fringe benefit" 2. Ignorance of the potential benefits
of the task of collecting enough of A.S., for them.
information to say something more or

981

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

3. Fear. to get the information the next best way


- and that's by asking you to estimate
4. Skepticism. it. Can you suggest anything better?"

Buy the Latest Conference Proceedings and Find Latest Computer Performance Management 'How To' for All Platforms at www.cmg.org
Once you understand the motives, and the .•• 1 don't have the time to work with
commonest excuses for not cooperating, you"
you can explore the inventory of
techniques for dissipating barriers and Answer: "First of all, here in the early
enlisting help. Here's a starter set, stages, I'm only asking for a 1 ittle of
together with some comments that should your time - an hour perhaps. If there's
help you map a strategy. The list is really no time to meet, how about lunch,
Join over 14,000 peers - subscribe to free CMG publication, MeasureIT(tm), at www.cmg.org/subscribe

surely incomplete, and you're invited to or a drink after work - on me!


extend it from your own experiences.
[Note: Please let the author know about "But think of it this way: the
any good ones you come up with!] See if information you provide now may help
you can guess which motive(s) go with uncover a design flaw that would have to
which excuses. be reprogrammed later. Then you'd have
to come up with even more time, nO
"I can't give you the information you're matter what. But it would be worse then:
asking for because ... the customer would be unhappy,
development costs would increase, and
... it's not my job / you're asking me to you'd be delayed from getting On to your
do YOUR job" next project. Frankly, you wouldn't look
too good.
Answer: "Right - in the narrowest sense.
your job is to translate specs into "Don't you think an hour's worth of
code, or to write specs, or whatever. insurance is worth it?"
But you're building a new system, and
only you know the details of the part .•• 1 don't understand what you're asking
you're handling. Sure, the system could for"
fail if it gives wrong answers, or
abends, or misses an important piece of Answer: "O.K., try this. Why not take a
functionality. But it could also fail if few minutes to tell me a little about
it runs too slowly, or takes too much the work you're doing on this new
resource, or costs too much. You playa application. That'll probably help me
key role in that too, because ... [fill relate what I'm looking for to things
in the way his/her function is you've probably already thought about
relevant]. And if it fails for and know pretty well.
performance or cost or any other reason,
we all pay the price. "I usually fi nd that people know more,
sometimes much more, than the simple,
"True, my job is to put together an summarized items I need. What I'm
assessment of the resource, performance looking for relates just to your part of
and cost picture for the application the overall application. [Give an
you're helping to build. I want to do my example of the sort of answer you're
job, but can't do it without a little looking for]. I'm not asking you to dot
help from you. every "i" and cross every "t".
"Look at it this way: I want to help "Of course, I'm a little new at this,
provide the right computer facilities to 1ike you, and if I ask for i nformat ion
run your application, just as a tailor that's not in your area, please tell me,
wants his suit to fit his customer and maybe even direct me to a better
perfectly. He and I both have to know source." [Note: Your A.S. modeling
something in advance to ensure a good software may help, by dividing the input
fit. He measures his customer, but by along functional lines, each "section"
the time I can measure a new asking for data which one person can
application, it's too late. So I've got normally provide.]

982

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

... 1 don't know it" "We'll be comi ng back to you duri ng


later development phases anyway, to ask
Answer: "You probably think that what if you want to revise any of your
I'm looking for is the exact answer - original estimates. But you can surely
the one the measurements-wTTl reveal help us by letting us know at any time

Buy the Latest Conference Proceedings and Find Latest Computer Performance Management 'How To' for All Platforms at www.cmg.org
when the system goes into production. if something important changes as a
But I'm not. All I'm asking for is your result of your work."
current best guess, or at least a
high/low range if you feel more ... it's out of my hands"
comfortable putting it that way. I only
ask that you give it a little thought, Answer: "Not really. Certainly there are
Join over 14,000 peers - subscribe to free CMG publication, MeasureIT(tm), at www.cmg.org/subscribe

and your best effort. some parts of the application that are
out of your hands, and others which you
"On my side, I'll recognize it as a have a say in. We're only asking for the
guess, and treat it as such. I'll latter. If we make a mistake, and ask
especially note where in the development you for parts that deal with another
life cycle it came from, and accord it area, perhaps you can suggest who we
the right degree of credibility. Most should talk to instead."
important, I'll be the buffer for anyone
who mistakes it for a hard and fast ... the high visibility of your reports
guarantee. You can count on me for that! compromises my ability to change my mind
later on"
"But someone has to do the early
guesswork, and if not you, then who?" Answer: "We'll do our best to tell
others what we've already told you: a
... it could change - it could be much guess is just a guess, and is subject to
bigger or smaller than what I think it change. We know right from the start
is right now" that it's an uphill battle to get
everyone to understand the difference
Answer: "Right - that's part of the between a guess and a guarantee.
game. Nobody expects your guesses to be Especially people who normally deal only
perfect - only to be your best shot. We with hard numbers like measurements or
understand that some of them may be contracts. But don't worry about having
pretty good, and others may change quite to deal with those people - that's our
a bit as analysis proceeds and the responsibility. and we'll live up to it.
application takes more definite form. Bet on it!"
That's why I'll even accept a range for
now. ... you'll use it to draw wrong
conclusions, or at any rate conclusions
"An estimate is just a place for us to I don't agree with, and then blame them
start. Once we get all the estimates on me"
together, we'll take a look at the
application as a whole. Usually when we Answer: "The conclusions we come to are
do this, some estimates turn out to be based on two kinds of inputs: the
more crit i cal than others - 1. e., if information we receive during our
they change it greatly influences the interviews, and some formal techniques
performance or cost of the application. (models) for processing that
Once we figure out which they are, we'll information. The modeling techniques
ask you to watch them, and let us know have been widely tested, and there's
if, for any reason, you want to raise or plenty of documentation to show that
lower your original estimates. they're really solid.
"Our purpose is to set up an early "On the other hand, the estimates come
warniQ9 framework. If a critica from a particular phase of development,
estimate changes, we want to know about so we know from experience roughly how
it, so that we can evaluate our mid- much credibility they deserve. Our
course correction options, and choose a conclusions have about the same degree
reasonable new course. of credibility. We don't represent them

983

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

as the absolute truth, any more than you and which are efficient. I see which
make the same claim for your estimates." transactions will cost a lot (and why),
which ones will run quickly, and which
... it's not worth it, because ones will be slow .
conclusions based on vague or uncertain

Buy the Latest Conference Proceedings and Find Latest Computer Performance Management 'How To' for All Platforms at www.cmg.org
data are worthless, no matter how "Now, here's the payoff. If we choose
carefully you analyze that data" to, we can keep everything fixed, except
your estimates. Those can be changed,
Answer: "Not really - it depends on your repeatedly, and the analysis can be
notion of value. Sometimes, knowing just re-run for each change - leading to new
a little more than nothing, but knowing performance, resource and cost results.
Join over 14,000 peers - subscribe to free CMG publication, MeasureIT(tm), at www.cmg.org/subscribe

it very early, has lots of value. Different design choices, different


estimates, different results. That means
"Here's an experience we heard of from you have the chance to evaluate four own
another company. It seemed like one of design choices, from the point 0 view
their new applications was going to or-their effect on the application as a
require a major CPU upgrade in a year whole. Isn't that something you'd like
when no upgrades were budgeted. By to know beforehand?"
working on the problem early on, using
A.S., they discovered that the right "In fact, isn't that something you've
question was 'How \many 3090-400's do we always wanted to know?
need?' rather than 'Do we need one at
all?' That's pretty valuable, don't you "What I'm saying is, once things get
think? rolling, A.S. techniques offer a kind of
"what if?" laboratory for you to check
"But take it a step further. If that out a huge number of design choices
were our situation, we'd have lots of before you actually make your decisions.
alternatives: negotiating to postpone In the end, that can only make you look
the application, or to introduce its good, because your systems will not only
functionality slowly (some this year, deliver the functions the end user asked
some next). Obviously, we can sometimes for, they'll run quickly, and at the
also negotiate for a budget variance, most reasonable possible cost. In the
and so on. All of these may be hard process, it'll make all of us look good.
choices, but it clearly pays to do the
analysis early, if only to make sure "All for the price of a few estimates
that no alternatives are missed, and now and again!
that people have time to get used to the
idea." "By the way, if you want to learn how to
use the modeling software yourself, I
... there's nothing in it for me" can set that up for you too."
Answer: "This is one of the most common,
and tragic misunderstandings. There's OVERCOMING THE OBJECTIONS: LIGHT AT LAST
lots in it for you - let me illustrate.
The objections in the previous section,
"As things stand now, I'm asking you to and the suggestions for overcoming them,
estimate certain parts of the are really just the start of successful
application. For example, if you're an A.S. Objections provide opportunities to
application designer, I'm asking you for sell the concept and benefits of A.S.,
the major transaction types, the number but they're not the only times to gain
and kind of system and DB calls they points. On the positive side, there are
make, and few other items. I pool your a host of specific ideas to get the
information with everyone else's, put it organization ready for integration of
into a performance model, and produce A.S. into all phases of the development
some very interesting and valuable life cycle. In this section, we discuss
numbers: CPU and I/O usage, a picture of a few of the most promising ones.
which transactions are resource hogs,

984

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

* Sell senior management on the (End user/application design) The


benefits of A.S., up front. ability to evaluate tradeoffs between
more/less functionality and the final
Clearly, nothing helps your cause so operational cost of the application -

Buy the Latest Conference Proceedings and Find Latest Computer Performance Management 'How To' for All Platforms at www.cmg.org
immediately as senior management clearly, simply and directly.
support. Development and Technical
Services executives have the power to In particular, the possibility to
adjust job definitions, and a mandate discover opportunities to increase
from them to incorporate A.S. into the functionality at little or no
Systems Development Life Cycle is the operational cost or performance
surest waY to make it happen. But since impact.
Join over 14,000 peers - subscribe to free CMG publication, MeasureIT(tm), at www.cmg.org/subscribe

A.S. usually represents a change to the


old ways of developing systems, you have (Operations) Discovery of
to prove your point - it's not opportunities to improve work
automatic. scheduling. Also, a better
understanding of spare capacity -
To get you started, here's a brief look which comes in handy when a sudden,
at some of the benefits, and costs, of unexpected, business-driven "swell"
A.S .. Dn the benefit side: appears in the workload, as it so
often does.
(End user/financial analysis)
Cost/benefit analyses can be based on To be fair, one has to explain that A.S.
a much more accurate estimate of has costs as well - but they're really
operational costs (versus, for small compared to the benefits:
example, extrapolating from current
applications). A coordinator's time, first for his
own training, and then ongoing time
(Capacity Planning/financial to carry out the activity. His time
analysis) More accurate capacity will be spent sporadically, as each
planning, because new applications project phase nears completion, in
are sized more precisely, at less order to coordinate a recap of the
risk, earlier. Growth drives the application as of the end of that
capacity plan, and new applications phase.
usually account for a major share of
growth. So accurate sizing means If major surprises appear, he'll have
fewer capacity surprises, and more to spend additional time to analyze
predictable budgets. Everyone likes possible responses, but this isn't
that. expected to happen too often.
(Financial analysis) More stable Obviously, the coordinator will also
chargeback rates - mid-year changes spend some time training others,
to the hardware plan, or to workload especially those who provide his
forecasts, can mean large variations estimates during the development
in the chargeback rates from month to process.
month, or quarter to quarter, in
order to end the year with cost Other people's time - in general, not
recovery right at 100%. much of it, unless they adopt
modeling as a personal technique.
(Application/data design) Discovery They'll spend a little more time but,
of tuning opportunities, i.e., as explained earlier, it will be well
changes which result in (sometimes worth it in the long run, since
drastically) reduced resource applications will tend to be
requirements: considerably more efficient and less
expensive.
- CPU and I/O usage
DASD space Machine resources again, these
- Network usage should be quite modest.

985

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

Software tools, in particular, tools them a while to realize it!


for analytic modeling of new
application performance. * Recognize that different people speak
different languages (jargon) and that
* Start with small steps, and increment there's an enormous advantage to

Buy the Latest Conference Proceedings and Find Latest Computer Performance Management 'How To' for All Platforms at www.cmg.org
slowly. localize benefits, and being a polyglot.
demonstrate them early, and often.
Remember that each participant is
For most shops, A.S. is a new, untried familiar with only a small part of the
technology, and modest results should be input - all the other inputs are of much
sought at first. In analyzing design less interest, and may even be
Join over 14,000 peers - subscribe to free CMG publication, MeasureIT(tm), at www.cmg.org/subscribe

alternatives, stick to those which are inscrutable. When you review a modeling
most easily and robustly modeled, and study, make sure to confine your
leave sophisticated modeling until attention to those inputs and outputs
later. However, it's absolutely your client can understand and has
essential that each participant sees, as interest in. It's a bit of a job, but
soon as possible, at least one you can do it.
comparison of alternative inputs he's
fami 1i ar with. * Recognize that nobody bats 1,000.
In the early phases, participants may You'll make mistakes, and so will the
well ask if they can use the modeling people who provide estimates for you. In
software on their own. Be prepared to the right eovironment, you'll all learn
tell them what's involved, and make sure from your mistakes - there's nothing
the answer is "yes". tragic about a mistake as long as you
catch it in time, and respond sensibly.
You should observe that people will You get better at anything you practice,
incorporate new technologies if they even estimating.
perceive them to be in their own best
interests, whether those technologies And if your senior management expects
are "part of their job" or not. In the perfection right from the start - well,
long haul, the benefitS will sell that's a mistake too!
themselves if they're presented right.
* Stay open to the wide range of
* Develop a vision of how available choices when it looks like
responsibility should be divided for an application is going to be too
A.S. to work best. big, expensive, or slow.
As the A.S. coordinator, you'll probably It's your job to present the problem,
begin by gathering raw estimates for tbe the options, and your evaluation of
new applications. Then you'll analyze them, NOT to decide between them. There
them with your modeling software, and are lots of toes you might step on in
use the results for your capacity the process. Managers are paid precisely
planning studies. That's not bad for to make tradeoffs, keeping all the
starters, but sooner or later the players reasonably content in the
applications are going to overwhelm you: process.
there are going to be too many for one
person to chase. * Circulate a ·composite estimates·
document, as first-level feedback to
On the other hand, skeptics can become all your estimators.
bel ievers. If you successfully sell the
idea of individual benefits for the Such a document, a complete listing of
application and data designers, they'll the "current corporate view of the
almost certainly ask to pick up application", is simply invaluable. It
responsibility for modeling their encourages cross-checking of
individual applications. And strictly assumptions, and increases mutual
speaking, that's the right place for the awareness among groups who may not
responsibility to reside. It just takes communicate enough. It strongly

986

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

encourages "second thought" revisions, "victim of change" attitude so common


and tends to cause views to converge, elsewhere. And this control over change
and inconsistencies to be aired and derives from a balanced division of
resolved. labor - with the development staff
handling the part of the sizing process

Buy the Latest Conference Proceedings and Find Latest Computer Performance Management 'How To' for All Platforms at www.cmg.org
* Don't be afraid to use "pie in the it alone can direct, while technical
sky" estimates (based, for example, services pools development inputs into a
on "industry standard transaction coherent, and well-founded capacity
profiles") when absolutely no data is plan.
forthcoming.
When mid-course corrections are needed,
Join over 14,000 peers - subscribe to free CMG publication, MeasureIT(tm), at www.cmg.org/subscribe

Sooner or later, despite your best they are negotiated sensibly, without
efforts, you'll encounter some groups trauma. If an appl ication is larger than
who simply fail to provide you with the expected, for instance, there's probably
needed estimates. Don't despair - a good reason for it. That fact just has
there's a technique that works most of to be understood and accommodated.
the time. Using your best judgement (and
any industry-wide averages you may know Finally, there's a change in the way
about), take a wi ld guess at the design decisions are made. Before A.S.,
estimates. such decisions are made in a vacuum.
Initially, designers tend to view models
Include your wild guess in the composite as threats, precisely because they have
estimates document, and label it for the power to reveal weak or flawed
what it is: a placeholder until some decisions after the decisions are "cast
better estimates come along. Given the in stone". Simply reversing the
visibility of your wild guesses, you'll procedure, i.e., running models before
be amazed how quickly the recalcitrant decisions are made, and in support of
group comes around. them, turns the whole situation around.
rlaws are spotted quickly, and avoided -
so decisions that "go public" have
CONCLUSION: THEY ALL LIVED HAPPILY EVER already been tempered by the fire of
AFTER analysis, and are usually far better for
that.
A few features characterize the
organization which successfully embraces All of this is possible, and true. It
A.S. Most of them are suggested by the suggests that integrating A.S. into the
techniques above. There is a working development process is well within the
relationship between the development reach of many, perhaps most,
staff and technical services which organizations.
transcends the traditional enmity
between these two groups. In particular,
the A.S. function serves as an example References (available from BGS Systems)
that, given the right respect for each
others jargon, meaningful and fruitful
communication is possible. [1] CRYSTAL Performance Evaluator User's
Manual, Document #90010-00
There is a self-assured sense of control [2) Rimicci, C., Predicting Application
over change, a sharp contrast to the Performance, (unpub1i shed note).

987

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

You might also like