The Structure and Value of Modularity in Software Design

1
The Structure and Value of Modularity in Software
Design

• Written by
• Kevin J. Sullivan, William G. Griswold, Yuanfang
  Cai
• Presented by
• Meltem Yildirim

CmpE-550 / 7.12.2005
2
Outline

• General Terms
• Modularity
• Information Hiding
• Value of Modularity
• Motivation of the Paper
• Baldwin Clarks Theory
• DSM
• NOV
• Contribution of the Paper
• Conclusion

3
General Terms

• Modularity the characteristic of a system that
  has been divided into smaller subsystems which
  interact with eachother
• Information Hiding hiding details of an object
  or function so that each module has the
  information just needed by it
• Parnas introduced information hiding in 1972
• case study comparing the changeability of two
  versions of KWIC (Key Words in Context)
• Hide the design decisions, which are most likely
  to change, from other parts of the system
• Value of Modularization the benefits obtained by
  exercising a specific modularization solution

4
Motivation

• We need a model to assess the relative values of
  alternative modularizations for the same system
• Which of the available modularizations is the
  best for a given system?
• Augment Parnass work
• Explore the potential of Baldwin Clarks Theory
  (2000)
• Design is a value seeking process
• DSM (Design Structure Matrix)
• NOV (Net Options Value)

5
Design Structure Matrix (DSM)
Design space to be searched for valuable designs
Design Parameters A, B, C
Design Parameters Data structures Algorithms User
interface look-and-feel Security aspects Power
Usage
For modularity, dependencies across proto-modules
must be broken by splitting
B depends on A
Other operators Substitution Augmentation Exclusi
on Inversion Porting
6
Net Options Value (NOV)

• Mathematical model for evaluating designs
• General expression for NOV of a modular design
• V S0 NOV1 NOV2 NOVn (1)
• NOVi maxki sini1/2 Q(ki) Ci(ni)ki - Zi (2)
• Zi Sj-sees-icinj (3)

Total Value of the System
NOV of the ith Module
Cost incurred in ki experiments
Redesign Cost
Value of the Unmodularized System
Visibility Cost of Replacing the ith Module
Expected Benefit
Complexity of the jth module
all j modules that see the ith module
7
Contribution

• Evaluate the applicability of Baldwin Clarks
  theory to inform software design
• EDSM extended DSM to also model environment
  variables which cannot be controlled by the
  designer
• Case study
• DSM-based analysis of KWIC
• EDSM-based analysis of KWIC
• NOV-based analysis of KWIC

• Computer Configuration
• (device capacity, speed)
• Corpus Properties
• (input size, language)
• User Profile
• (experienced or not, interactive or offline)

8
EDSM for strawman modularization
EDSM for proto-modular design
EDSM for information-hiding
modularization
9
NOV Analysis of KWIC

• V for strawman design 0.26
• V for proto-modular design 1
• V for information-hiding design 1.56

x 6
10
Conclusion

• EDSMs have the potential to
• model the software and its environment
• capture Parnas information hiding criterion
• Compare different modularizations
• NOV formula provides a quantitative value for
  good modular design
• Information-hiding design is superior to strawman
  design

11
Questions?
12
My Opinion

• The process of splitting modules so that they do
  not affect each other makes sense
• Case study KWIC
• It is not clear what the modules are, how they
  are chosen and how they are manipulated
• Scalability is questionable, may be hard to
  utilize in practice because commercial
  applications are far more complex than KWIC
• The use of Hierarchy Diagrams in designing
  large systems may be beneficial