Cost Effectiveness of Software Reuse Alternatives

© 2008, Dr. Amir Tomer
ISEC 2008, Hyderabad, India
Cost-effectiveness of Software Reuse Alternatives 1

Education
B.Sc., M.Sc. – Computer Science, Technion, Haifa, Israel
Ph.D. – Computing, Imperial College, London, UK
Industrial Experience
1982-2006, RAFAEL – Advanced Defense Systems, Israel
Software developer, software manager
Systems Engineer
Director of Systems and Software Engineering Processes
Academic Experience
Senior lecturer, Software Engineering
Academic advisor, Software courses, External Studies
Certifications
Certified Software Quality Engineer (CSQE) by ASQ
Authorized CMMI instructor by SEI

• Well known facts
– Software reuse can save costs
– Generic software development costs more than dedicated
software development
• However...
– Can the saving be estimated with high confidence?
– Can the ROI of generic development be precisely predicted?

• This tutorial introduces a model for evaluating the
relative cost-effectiveness of software reuse
alternatives
• The model is
– Formally defined
– Approved by industry
– Effective
– Easy to implement
and...
– Fun!

 Once upon a time, on a project kickoff meeting...
 The “Reuse” issue from the cost perspective
 Reuse-based development – operations
 Elements of reuse costs
 Reuse scenarios
 Costing policies
 Industrial case study
 Wrap up

GoldenEgg Project
Kick-off Meeting
Participants:
Project Manager
System Engineer
Software Manager
DB Architect
Network Engineer
Division R&D Manager
Sales Manager

It was proposed to reuse
the Backup & Recovery
module from one of the
previous projects.
What do you think?
Great!
Brilliant
idea!
Direct saving of
1500 hours!
Let’s go for it!
I always support
reuse!
It should shorten
the schedule too!
1 Backup & Recovery (BR)
Development Cost Estimation
Network & Server 600 p/h
Application Software 600 p/h
Integration & Test 200 p/h
User Documentation 100 p/h
___________________________________
Total 1500 p/h

Wait, wait...
We first need to scan through all
past projects and then analyze
which BR module fits best.
Investing 50 p/h in this research
should be very cost-effective
2

New communication standards have been introduced, and the
network design must be amended accordingly. It should take
100 p/h.
And, by the way, the DB server has been upgraded too, so in
any case slight design modifications are necessary. These
should take not more than another 150 p/h, but this should
serve the future projects as well.
3

An upgraded server requires a
few software changes.
Assuming the DB schema
persists then my estimation
does not exceed 250 p/h.
Actually we did
developed an improved
schema, but you can
have it for free.
But if the schema
changes, then the reuse
is meaningless! The
entire software must be
re-developed.
So, if you go for re-development
let’s look ahead
to the future projects. With
800 p/h we can develop a
generic DB architecture to
support future needs.
6
7
5
4
8
To my experience, a generic
architecture will double the
cost of software development,
although the chances for
reuse increase. However, in
any case, each project should
build its own interface,
investing 100 p/h.

A generic software is more
complex and may double
integration and test effort,
although it will be done once for
future projects.
I suggest that R&D will support
the generic development
9

I am not supporting anything
without proven ROI. In order to
10
allocate R&D budget I need to know
how many future projects are
expected to use this generic module.
Do we have any such anticipation?
We currently negotiate 10
prospective clients. Assuming
50% success gives 5 projects.
Adding the current one makes a
total of 6 reuses.
11

The existing solution is applicable to 2-3
projects, at the most. Afterwards a new
DB architecture will be needed anyway. If
we do not go for a generic solution every
future project will have to develop their
own module separately.
12

So, what do we decide?
?
?
?
?
?
?
“Traditional” development 1500 p/h
Seek a reusable module 50 p/h
New comm. standards 100 p/h
Network changes - new server 150 p/h
SW changes for new server 250 p/h
Generic DB schema 800 p/h
SW for generic schema 1200 p/h
DB interface – each project 100 p/h
Generic module IV&V 400 p/h
Reuse anticipation = 6

• What is the best decision for the project manager, if R&D
budget is not granted?
• Same situation, but the PM is responsible for all future
projects
• What is the cost of the first 3 projects in each of the
alternatives?
• How many future projects should be contracted in order to
convince R&D to budget generic development?
Would you like to make a guess?
By the end of this tutorial we will be able to answer
these questions precisely...

 Reuse scenarios
 Wrap up

• Question 1: What can be reused?
• Answer 1: Any kind of software artifacts/components
– Requirements
– Features / specifications
– Designs
– Source code
– Run-time components (EXE, DLL, COM objects etc.)
– Documentation (e.g. manuals, help files)
– Test cases / Test procedures
– Commercial off-the-Shelf (COTS)

• Question 2: What is affected?
• Answer 2: Could be anything
– Nothing!
• Is this possible at all?
– The reused component is adapted to fit the new
environment
• What is the amount of modification needed?
• What “damage” is caused to the component?
– The environment is adapted for the component to fit in
• What is the effect on other components in the same
environment?
Huge
Differences!

• Question 3: How components are being reused?
• Answer 3: Two conventional forms
– Black-box reuse
• The component is used unchanged (as is)
– White-box reuse
• The component undergoes modification


100%
development/
modification
effort
0%
no reuse
black-box
reuse
white-box
reuse

Informal definitions, just for the common understanding of what follows
• Software Artifact
– Any work product resulted from a software development activity
• Software Component
– A piece of software or a part of a software system, including all the
artifacts associated with it
• Software Asset
– Any software component or artifact residing in library, accessible to
relevant stakeholders
• Assets may be either private (owned by a single project) or public (accessible
to one or more organizational units)
• Core Asset
– A public asset which is classified as an important reusable building block

• Costs – from which viewpoint?
– We look at cost from the enterprise point of view
• Over all projects
• Over all NRE (R&D)
• Over all supporting infrastructure
• The cost of any delivered product comprises the
following ingredients
– The cost of the specific product development
– The cost of reusable assets incorporated into the product
– The cost of the infrastructure enabling asset reuse

• Import of components into the product
– Searching and locating reusable assets
• From inside or outside the organization
– COTS procurement
– Outsourcing component development
• Implementation of product components
– Requirements, analysis, design, code and unit test
– Applying modifications/customizations to imported components +
regression test
– Applying modifications to other components, in order to enable the
incorporation of imported components
• Product integration, verification and validation (IV&V)
– Design reviews
– Partial and final integrations
– System test, acceptance test, validation etc.
– ATE development / procurement

• Import of core assets
– Searching and locating reusable assets
• From inside or outside the organization
– COTS procurement
– Outsourcing component development
• Implementation of core assets (generic/reusable components)
– Requirements, analysis, design, code and unit test
– Applying modifications/customizations to imported components +
regression test
– Applying modifications to other components, in order to enable the
incorporation of imported components

• Asset Repository Establishment & Maintenance
– Storage and communication
– Database services
– Catalog services
• Asset Repository Population
– Domain Analysis
• Mapping the organization product lines
• Categorization and specification of reusable assets
– Storing and cataloging identified assets

• From your knowledge and experience, recall a case of
reusing a software component, and then answer:
Q: What was the component in focus?
A: __________________________________________
Q: What activities were performed?
A: __________________________________________
Q: What is the estimated cost (by means of effort, money or both )
A: ____________________________________________

• Who covers the costs of core asset development?
• How are these costs shared among the products using
the assets?
and most importantly
• Is it always cost-effective,
and if not
• When is it? And under what conditions?
In order to answer these questions a better breakdown of
the costs is needed...

 Reuse scenarios
 Wrap up

© 1999-2008 , ד"ר עמיר תומר
2-dimensional evolution tree:
The life cycle of a software component [Schach & Tomer, 2000]
recoding
rerequirements
reanalysis
redesign
recoding
redesign
C recoding 1
Ver. 1
R2
S2
D2
C2
Ver. 2
D3
C3
Ver. 3
f
R0
S0
D0
C0
Ver. 0
Require-ments
Analysis
Design
Code
Development Axis
Maintenance Axis

3-dimensional evolution tree:
The life cycle of a software product-line [Schach & Tomer, 2000]
diversity
Reuse
Core asset repository
Development
progress
Maintenance
effort
• Evolution occurs by the
“moves” of artifacts along
the 3 axes:
– Development
• Engineering
• Reverse-engineering
– Maintenance
• Modification
• Roll-back
– Reuse
• Mining
• Acquisition
Mining an
existing asset
Acquiring a
core asset

Repository
Assets
Private
Assets
• Artifacts that are cataloged in the core-asset repository
– Available for acquisition by specific products
– Available for modification, only within the environment of the repository
– Include metadata (classification data and attributes)
• Artifacts that are contained within a specific product
– Available for mining and cataloging into the repository
– Available for private modification, only within the environment of the
same specific product

Repository
Assets
Private
Assets
Transformation Operations
Transition Operations
Transformation Operations
Development, Maintenance
Reuse

• Repository asset transformation
– NR: New for Reuse
• Constructing a new repository asset R' from scratch
– It is expected that the asset will be developed in conformance
with applicable standards that make it effectively reusable
– AR: Adaptation for Reuse
• Modifying an existing repository asset R, resulting in another
reusable repository asset R‘
– Both R and R' must reside in the repository
Repository
New for Reuse [NR]
R R’ Assets Adaptation for
Reuse [AR]

• Private asset transformation
– ND: New Development
• Constructing a new private asset P' from scratch
– WB: White-Box reuse
• Modifying an existing private asset P into another private
asset P' within the same product
– Both P and P' must reside in the product library, as revisions of
the same artifact
White-Box
Reuse [WB] Private
P P’
Assets
New Development [ND]

• Transition into Repository
– MC: Mining and Cataloging
• Identifying and acquiring an existing private asset P, from a
certain product, and then storing and cataloging it formally,
as a repository asset R
– XA: eXternal Acquisition
• Acquiring an asset from some external source and cataloging
it as a repository asset R. This is the case with COTS
artifacts
eXternal Acquisition [XA]
Repository
Assets
Mining & Cataloging [MC]
Private
R
P Assets

• Transition into Product
– BB: Black-Box reuse
• Acquiring a copy of a repository asset R “as is” (i.e. with no
modifications) for a specific product, as a private asset P
– CP: “Copy and Paste”
• Acquiring a copy of a private asset P for a specific product
– The source asset is not cataloged in the repository
– Awareness of its existence is based on personal knowledge
P
Repository
Assets
Private
Assets
R
Black-Box reuse [BB]
“Copy & Paste” [CP]

New for Reuse [NR]
R R’
P P’
Cataloged asset
Acquisition (CA)
White-Box
Reuse [WB]
Transformation operations
Transition operations
Repository
Assets
“Copy & Paste” [CP]
Private
Assets
Adaptation for
Reuse [AR]
Black-Box
Reuse
[BB]
Mining &
Cataloging [MC]
eXternal (COTS)
Acquisition [XA]
target
component

• Refer to the reuse case you mentioned in Exercise 1 and
express the reuse activities in terms of one or more
elementary operations, specifying
– The source artifact(s) and their location(s)
________________________________________________
– The target artifact(s) and their location(s)
________________________________________________
– The elementary operation(s) used
_____________________ ______________________
_____________________ ______________________

 Reuse scenarios
 Wrap up

• Every operation is associated with costs
– Effort, in p/h
– Money
• Unlike common illusions, there are usually no free
operations!
• Costs may be classified by the concern of the body obtaining
the artifact
– Project costs
• The cost of any transformation occurring within the project
• The cost of any transition into the project
– Infrastructure costs
• The cost of any transformation occurring within the repository
• The cost of any transition into the repository

Let
OP be any elementary operation
S, T be source and target artifacts, respectively
Then
COP(S,T) denotes the cost incurred by obtaining T from S
by means of OP
• Examples
1. CAR(q.c, q.uml)
• The effort invested in reverse-engineering
2. CCA(q.uml, q.java)
• The price of a java code-generation license

• Part of the elementary operations result in new artifacts
(from scratch)
• These operations are equivalent to other operations, where
the source artifact is null (denoted “f”)
• Equivalent operations
– New for Reuse  Adaptation for Reuse (from scratch)
• CNR(R’) = CAR(f, R’)
– New Development  White-Box reuse (from scratch)
• CND(P’) = CWB(f, P’)
– eXternal Acquisition  Mining and Cataloging (from the outside)
• CXA(R) = CMC(f, R)
– “Copy & Paste”  Catalog Acquisition (of the empty artifact)
• CCP(P) = CCA(f, P)

CNR(R’) = CAR(f,R’)
f
R R’
P P’
Repository
Assets
Private
Assets
CAR(R,R’)
CMC(P,R)
CCA(R,P) CBB(R’,P’)
CND(P’) = CWB(f,P’)
f
f
CMC(P’,R’)
CCP(P) = CCA(f,P)
CWB(P,P’)
f
CXA(R) = CMC(f,R)
Infrastructure costs:
• Obtain Repository Assets
• Modify Repository Assets
Project costs:
• Obtain Private Assets
• Modify Private Assets

• When coming to evaluate alternatives, cost accumulation
may combine actual costs with estimated costs
– Actual costs: Measured in past activities
– Estimated costs: Predicted according to knowledge and
experience
• For example, consider a system integrated by two
components: A and B. The alternatives are:
Alternative A B Interface
1 Buy Buy Develop
2 Buy Develop -
3 Develop Buy -
4 Develop Develop -
Actual Costs
Estimated Costs

Cost Notation Measurement Estimation
Adaptation
for Reuse
CAR(R,R’) The actual cost invested in
turning the repository asset
R into a reusable repository
asset R’
The estimated cost required
for turning the repository
asset R into a reusable
repository asset R’
New for
Reuse
CNR(R’) =
CAR(f,R’)
The actual cost invested in
building the reusable
repository asset R’ from
scratch
for building the reusable
repository asset R’ from
scratch
White-Box
Reuse
CWB(P,P’) The actual cost invested in
turning the private asset P
into a specific private asset
P’
for turning the private asset
P into a specific private asset
P’
New
Development
CND(P’) =
CWB(f,P’)
building the specific private
asset P’ from scratch
for building the specific
private asset P’ from scratch

Cost Notation Measurement Estimation
Mining and
Cataloging
CMC(P,R) The actual cost invested in
locating the private asset P
and cataloging it as a
for locating the private asset
P and cataloging it as a
Cataloged
assets
Acquisition
CCA(R,P) The actual cost invested in
obtaining a copy of the
repository asset R as a
specific private asset P, for
further modifications
for obtaining a copy of the
specific private asset P, for
further modifications
eXternal
Acquisition
(COTS)
CXA(R) =
CMC(f,R)
The actual cost of purchasing
a copy of R as a repository
asset
The estimated cost of
purchasing a copy of R as a
repository asset
Copy and
Paste
CCP(P) =
CCA(f,P)
obtaining a copy of P as a
specific private asset
for obtaining a copy of P as a
specific private asset
Black-Box
Reuse
CBB(R’,P’) The actual cost invested in
obtaining a copy of the
specific private asset P “as is”
for obtaining a copy of the
specific private asset P “as is”

 Reuse scenarios
 Wrap up

• An asset is subject to a single elementary operation at
one time:
– Transformation
• Modifying the asset within its current location
– Transition
• Moving an unchanged copy of the asset from one location to
another

  

old C project
queue.c
new C++ project
queue.cpp
new Java project
queue.java
S/W Dept.
A C-language queue-handling
package from
an old project is
reverse-engineered into
UML design
The design is
implemented in C++
for a new project
queue.uml
The design is
implemented in Java
for a new project

• Scenario = a sequence of elementary operations
– Following the “one-move-at-a-time” principle
• The example scenario
1. MC: q.c is mined from the C project and cataloged in the repository
2. AR: q.c is adapted for reuse (reverse-engineered) into q.uml
3. CA: q.uml is acquired from the repository into the C++ project
4. WB: q.uml is privately implemented in the C++ project, resulting in q.cpp
5. CA: q.uml is acquired from the repository into the Java project
6. WB: q.uml is privately implemented in the Java project, resulting in q.java
Repository
C project
C++ project
Java project
1 [MC]
q.c
q.c
2 [AR]
q.uml
3 [CA]
q.uml
5 [CA]
q.uml
4 [WB]
q.cpp
6 [WB]
q.java

• Although the variety of reuse-based development scenarios is
unlimited, a few scenarios are more widely applied
• In the following we will investigate 5 typical scenarios
– Pure Development
• Traditional development from scratch, with no reuse at all
• This is a reference scenario for cost-effectiveness evaluation
– Opportunistic Reuse
• The most commonly used scenario, usually provides an instant
opportunity to save some development effort
– Controlled Reuse
• Sharing of reusable components in the organization is controlled, even
without developing generic core asset
– Systematic Reuse – new
• Reusable core assets are developed from scratch for standard use
across the organization
– Systematic Reuse – adapted
• Provisionally reusable assets are mined from past projects and then
being modified into core assets for reuse across the organization

• Description
– Each project i develops its own component Ti from scratch
• No reuse in effect
• Scenario sequence
– Single move, repeated at each project
• New Development [ND]
This scenario serves as a reference to all other
scenarios
Private
Assets
f T1,…,Tn

• Measured Cost
– The total cost is the accumulated costs of the development
effort over all the projects
CPD = n(CND(Ti)) = n(CWB(f,Ti))
• Estimated Cost
– Assuming T1  T2  ...  Tn  T
CPD  n · CWB(f,T)
Private
Assets
n(CWB(f,Ti))
f T1,…,Tn

Private
Assets
White-Box reuse [WB]
T1,…,Tn
S1,...,Sn
Copy & Paste [CP]
f
• Description
– Each projects i independently obtains an existing source
component Si and modifies it into a private component Ti by
means of White-Box reuse
• Scenario Sequence
– Two moves, repeated at each project
• Copy & Paste
• White-Box Reuse

Private
Assets
T1,…,Tn
n(CWB(Si,Ti))
S1,...,Sn
n(CCA(f,Si))
f
• Measured Cost
– The total cost combines the effort invested by each project in
searching and copying Si with the effort invested modifying Si
into Ti
COR = n(CCP(Si) + CWB(Si,Ti)) = n(CCA(f,Si) + CWB(Si,Ti))
• Estimated Cost
– Assuming S1  S2  ...  Sn  S and T1  T2  ...  Tn  T
COR  n · (CCA(f,S) + CWB(S,T))

• Description
– Projects are not allowed to freely obtain assets from other
projects
– Assets from past projects are cataloged in a searchable
repository
• By initiated mining
• By volunteered/embraced/rewarded contribution
• By organizing all private project libraries under a common
repository (organizational intranet)
– Assets remain in the repository in their original form
– Assets are available for Catalog Acquisition [CA] to all
projects

• Scenario Sequence
– Three moves
• Mining (once only!)a source asset S and Cataloging it as a
repository asset R
• Catalog Acquisition of a copy of R into a source asset Si, for each
project i
• Obtaining a private target asset Ti from Si in each project i
R Repository
Assets
Mining & Cataloging [MC] Catalog Acquisition [CA]
White-Box reuse [WB]
Private
Assets
Private
Assets S1,...,Sn T1,...,Tn
S

• Measured Cost
CCR = CMC (S,R) + n(CCA(R,Si)) + n(CWB(Si,Ti))
• Estimated Cost
CCR  CMC (S,R) + n · (CCA(R,Si) + CWB(Si,Ti))
R Repository
Assets
CMC(S,R) n(CCA(R,Si))
n(CWB(Si,Ti))
Private
Assets
Private
Assets S1,...,Sn T1,...,Tn
S

• Description
– A central unit at the organization (or product-line) level is responsible
for developing core assets and making them available to projects
• new: Develop core assets from scratch
• adapted: Develop core assets by adapting assets which were mined from past
projects
– Usually these assets are generic
• Satisfy various needs
• Fit into various environments
– These core assets are reused in projects “as is”
• Black-Box reuse
– In many cases the use of such assets is subject to organizational
discipline
• No project is allowed to develop or acquire such assets independently
• The central core-asset group support the improvement of these assets
according to changing project needs

• Sequence
– new: two moves
• Develop (once!) a New core asset R for Reuse
• Black-Box reuse copies of R as target private asset Ti at each project i
– adapted: three moves
• Mine a reusable asset S from a past project and Catalog it as a core asset R’
• Adapt R’ for Reuse resulting in a core asset R
• Black-Box reuse copies of R as target private asset Ti at each project i
R Adapt for
R’
Repository
Assets
New for Reuse [NR]
Black-Box reuse [BB] Mine and Catalog [MC] Black-Box reuse [BB]
Private
Assets
Private
R
Reuse [AR]
T Assets 1,...,Tn
S
T1,...,Tn
60

• Measured Cost
CSRn = CAR(f,R) + n(CBB(R,Ti))
CSRa = CMC(S,R’) + CAR(R,R’) + n(CBB(R,Ti))
• Estimated Cost
CSRn  CAR(f,R) + n · CBB(R,Ti)
CSRa  CMC(S,R’) + CAR(R,R’) + n · CBB(R,Ti)
CARC (R’,R) AR(f,R)
R’
Repository
Assets
CMC(S,R’)
n(CBB(R,Ti))
Private
Assets
Private
T Assets 1,...,Tn
S
R
R
n(CBB(R,Ti))
T1,...,Tn

Well, let’s try the model
on the BR module...
Sure
Fine
Yes!
OK
Sounds good
I’m in
“Traditional” development 1500 p/h
Seek a reusable module 50 p/h
New comm. standards 100 p/h
Network changes - new server 150 p/h
SW changes for new server 250 p/h
Generic DB schema 800 p/h
SW for generic schema 1200 p/h
DB interface – each project 100 p/h
Generic module IV&V 400 p/h
Reuse anticipation = 6

S/W IV&V Doc. Total
N/W &
server
DB
schema
Elementary Cost Cost Details
0 250 250 200 100 800
Adapt a past module for
generic reuse in all projects
(same as CWB)
CAR
Adaptation for
Reuse
800 600 1200 400 100 3100
Develop a new generic module
from scratch
CNR New for Reuse
0 250 250 200 100 800
Modify a past module into a
specific module for a single
project
CWB White-Box Reuse
0 600 600 200 100 1500
Develop the module from
scratch for a single project
CND New Development
50
Search and obtain a past
module and make it available
to all projects (same as CCP)
CMC
Mining and
Cataloging
C Retrieve an asset from Catalog 0 CA
Cataloged asset
Acquisition
N/A
Purchase the module off-the-shelf
CXA
eXternal
Acquisition (COTS)
50
Search and obtain a past
module for the specific project
CCP Copy and Paste
100 100
Implement a driver for the
generic module at a project
CBB Black-Box Reuse

• Pure Development
– Each of the n projects develops its module independently
CPD = n · CND = 1500·n
• Opportunistic Reuse
– Each project independently obtains and modifies the module
COR = n · (CCP+CWB) = 850·n
• Controlled Reuse (applicable only for n≤3 !)
– First project obtains and modifies the module and the other use
it as is
CCR = CMC + CAR + n · CBB = 850 + 100·n
• Systematic Reuse (new)
– A brand new generic module is developed for the benefit of all
projects
CSR = CNR + n · CBB = 3100 + 100·n

10000
9000
8000
7000
6000
5000
4000
3000
2000
1000
0
1 2 3 4 5 6
Pure Development Opportunistic Reuse Controlled Reuse Systematic Reuse

• What is the best decision for the project manager, if R&D budget is not
granted?
– Opportunistic reuse – find something and adapt it
– Pure Development – forget about reuse
• The two alternatives are pretty close
– A further risk analysis may support final decision
10000
9000
8000
7000
6000
5000
4000
3000
2000
1000
0
1 2 3 4 5 6

• Same situation, but the PM is responsible for all future projects
– Controlled Reuse - develop one and keep it
• Remember: This is applicable only for the next 2-3 projects
– Systematic Reuse
• Becomes attractive for >4 projects, even without R&D support
10000
9000
8000
7000
6000
5000
4000
3000
2000
1000
0
1 2 3 4 5 6

• What is the cost of the first 3 projects in each of the
alternatives?
• How many future projects should be contracted in order to
convince R&D to budget generic development?
Can you answer by yourself?
Answers:
• CPD=4500, COR=2550, CCR=1150, CSR=3400
• >4

 Reuse scenarios
 Wrap up

• Many companies operate in a “Matrix Organization”
– Software assets are developed in a central unit
– Assets are delivered internally to projects
• See also “Framework for Software Product Line Practice” by the SEI
– http://www.sei.cmu.edu/productlines/framework.html
• How is each project charged for the core assets?
– Asset development
– Infrastructure costs

Infrastructure
Core asset
construction
Product
Costs construction
Operations
Domain
analysis
Reposi-tory
Store
and
catalog
Asset
devel-opment
Asset
acqui-sition
Product
IV&V
Asset
devel-opment
Asset
acqui-sition
C F F AR (R, R')
Trans-formation
C F F NR (R')
C F F WB (P, P')
C F F ND (P')
C F P CA (R, P)
Transition
C F CP (P)
C F F P P MC (P, R)
C F F P XA (R)
C F F F P BB (R', P')
F = Full: Cost of operation comprises the full direct cost of the cost component
P = Partial: Cost of operation comprises a relative part of the cost component

• Global R&D
– Projects get core assets for free
– R&D budget is derived globally from projects’ profit
• Unified costing, based on expected use
– NRE cost is distributed in advance over expected number of
reuses
– Each project is charged the relative part
• Variable costing – direct R&D
– Projects get R&D budget to support the development of their
private assets in a more generic fashion, for the benefit of
future projects
– Following projects “buy” assets, based on their level of
reusability

The Charge per project for the BR Module, n=1,...,10
3500
3000
2500
2000
1500
1000
500
0
1 2 3 4 5 6 7 8 9 10

• Risks
– Technology change
– Market / requirements stability
– Organizational structure stability (long term support)
• Schedule
– Time to Market constraints
• Quality
– The effect of regression testing
– The complexity of generic asset testing, compared to
dedicated assets

 Reuse scenarios
 Wrap up

• A Consortium of Israeli Industries
– Objective: Definition and evaluation of a unified software reuse
methodology
– Sponsorship: Chief Scientist, Ministry of Trade and Commerce
– Participating companies
• Defense Industry: RAFAEL, IAI, Tadiran Electronic Systems
• Civilian Industry: ECI Telecom, NICE, Orbotech, Creo (Kodak)
• 3-year project (6/2000 – 5/2003)
– Year 1: Literature review and methodology definition
– Year 2: Selective pilot at each industry and collection of unified
measurements
– Year 3: Result report and conclusions, based on the collected
measurements

• Data is based on a pilot of 7 assets in one company
– Red means actual measurements
– Black means estimations
1 2 3 4 5 6 7
Asset
Basic operation
Mining and Cataloging (MC) 110 110 110 110 110 110 110
Copy and Paste (CP) 110 110 110 110 110 110 110
New for Reuse (NR) 1000 2000 1000 1000 200 2000 2000
Adaptation for Reuse (AR) 500 500 400 400 50 400 1000
Black-Box reuse (BB) 100 200 150 100 30 400 10
New Development (ND) 700 1500 700 700 180 1500 1500
Catalog Acquisition (CA) 100 100 50 50 20 50 50
White-Box reuse (WB) 300 300 400 300 50 300 700
Number of reuses 2 3 3 2 3 3 4

• Cost of various Scenarios
– Red means actual measurements
– Black means estimations
1 2 3 4 5 6 7
Asset
Scenario
Systematic Reuse — adapted 810 1210 960 710 250 1710 1150
Systematic Reuse — new 1310 2710 1560 1310 400 3310 2150
Controlled Reuse 910 1310 1460 810 320 1160 3110
Opportunistic Reuse 820 1230 1530 820 480 1230 3240
Pure Development 1400 4500 2100 1400 540 4500 6000

• Cost-effectiveness ratio between scenarios for the 7 assets
1 2 3 4 5 6 7
Asset
Basic operation
Systematic Reuse vs.
Controlled Reuse 11% 8% 34% 12% 63%
Opportunistic Reuse 1% 2% 37% 13% 65%
Pure Development 42% 73% 54% 49% 81%
Controlled Reuse vs.
Systematic Reuse -28% 32%
Opportunistic Reuse 33% 6%
Pure Development 41% 74%
Number of reuses 2 3 3 2 3 3 4

3000
2500
2000
1500
1000
500
0
• Scenario comparison for asset # 4
1 2 3 4 5 6 7
Number of Anticipated Reuses
Cummulative Cost
Systematic Reuse - adapted Systematic Reuse - new Controlled Reuse
Opportunistic Reuse Pure Development

 Reuse scenarios
 Wrap up

• Reuse is not free
– But costs may be significantly reduced
• Reuse cannot be done without supporting infrastructure
– But a policy may be introduced for distributing the
overhead among projects
• There are always alternative reuse scenarios
– And they can be evaluated for their cost-effectiveness
• The number of expected reuses is a critical factor
– Established product-line strategy is required

• References
– Tomer, A. and Schach, S. R., “The Evolution Tree: A Maintenance-Oriented
Software Development Model”, Proc. CSMR’2000, pp. 209-214.
– Schach, S. R. and Tomer, A., “Development/Maintenance/Reuse: Software
Evolution in Product Lines”, Proc. SPLC1, Denver, USA, August, 2000, pp. 437-
450.
– Tomer, A. et al, “Evaluating Software Reuse Alternatives”, IEEE Transactions on
Software Engineering, Vol. 30, No. 9., September, 2004.
– Clements, P. & Northrop, L. , Software Product Lines: Practices and Patterns.
Addison-Wesley, 2002.
• Other Sources
– Morad, S. and Kuflik, T., “Conventional and open source software reuse at
Orbotech”, Proc. SwSTE’05, p. 8.
– Mohagheghi, P and Conradi, R., “Quality, productivity and economic benefits of
software reuse”, Empiricas Software Engineering, Vol. 12, No. 5 / October, 2007.
– ISO/IEC 12207:1995/Amd.1:2002(E) – Annex G

• Dr. Leah Goldin
– Golden Solution, Kfar Saba, Israel
• Dr. Tsvi Kuflik
– University of Haifa, Israel
• Dr. Esther Kimchi
– Ramat Hasharon, Israel
• Prof. Stephen R. Schach
– Vanderbilt University, Nashville, TN, USA

The complete article may be found in:
[Tomer et al, 2004]

Cost Effectiveness of Software Reuse Alternatives

More Related Content

What's hot (20)

Viewers also liked (20)

Similar to Cost Effectiveness of Software Reuse Alternatives (20)

More from Prof. Amir Tomer (11)

Recently uploaded (20)

Cost Effectiveness of Software Reuse Alternatives