SOFTWARE PRODUCT SIZE MEASUREMENT METHODS
A SYSTEMATIC MAPPING STUDY
Sohaib Shahid BajwaCigdem Gencel
Pekka Abrahamsson
Free University of Bozen – Bolzano
Bozen – Bolzano, Italy
OUTLINE• Introduction• Systematic Mapping Study (SMS)• Results & Discussion• Conclusions• Future Work
2
INTRODUCTION – 1/2• Size
• One of the basic attribute of software products
• Why size
• A key input for software effort and cost estimation
• To monitor project achievements
• Normalization of other measures when assessing quality (e.g. number of defects/SLOC)
• Major entitites (to measure software size)
• Software Requirement Specification (SRS)
• Software Design
• Software Code3
INTRODUCTION - 2/2 • Common approaches to software size measurements
• Identify key components of software product
• Measure each using rigorous measurement methods
• Issues related to software size measurement methods
• Immaturity and difficulties in applying methods
• Insufficient empirical/theoretical validation of measures
• Lack of standardization of methods
• During last decades
• New measures and methods were developed
• Improvements were made in exisiting ones
4
A SYSTEMATIC MAPPING STUDY (SMS) – 1/5 A SMS
Provides an overview of a research area Reduces the bias of researchers Well defined criteria to include/exclude primary
studies Repeatable
Classification rather than in depth analysis (Systematic Literature Review)
5
A SYSTEMATIC MAPPING STUDY (SMS) – 2/5 Research questions
Which software size measures and methods were developed in software engineering?
What is the current state of the art and trends in software sizing?
Search string (("software size" OR "software product size") AND
("measur*" OR "metric*")) Search strategy
Search databases Limited snowball sampling Main conference proceedings
6
A SYSTEMATIC MAPPING STUDY (SMS) - 3/5 Search databases
Database Name Total Found
Year
IEEE Xplore Digital Library 613 1977 – 2013ACM Digital Library 444 1947 – 2014Springer Link 489 1968 – 2014Total 1546
Database Name Total Number of Studies
Total (after inclusion criteria) 691IEEE Xplore Digital Library 276ACM Digital Library 241Springer Link 174
Total (after exclusion criteria) 515Total (after reading the full texts) 167
7
A SYSTEMATIC MAPPING STUDY (SMS) – 4/5 Snowball sampling (Studies become from 167 – 185) Inclusion from main conference proceedings ( studies
become from 185 – 208) Publication channel distributions
Publication types No. of PapersJournal 65 (31%)Conference/Workshops 124 (61%)
Book Chapter 10 (5%)Others 9 (4%)Total 208
8
A SYSTEMATIC MAPPING STUDY (SMS) – 5/5 Publication distribution per year
Increasing trend after 2008
9
RESULTS AND DISCUSSION – 1/5 79 Software product size measure/methods
Functional size classification
Software Product Size Measure/Method
Total
Functional Size 68Physical Size 11Total 79
Functional Size Classification Total
New/Novel 12Extended 33Tailored 23Total 68
10
RESULTS AND DISCUSSION – 2/5 Extended functional size measure/method
Tailored functional size measure/method
Functional Size Measure/Methods
Total
IFPUG 18COSMIC 9Others 6Total (Extended) 33
Functional Size Measure/Methods
Total
COSMIC 13IFPUG 6Others 4Total (Tailored) 23
11
RESULTS AND DISCUSSION – 3/5 Number of Functional Sizing Methods with Respect to
Years (New/Extended/Tailored) After 2008, most of the methods were either tailored or
extended
12
RESULTS AND DISCUSSIONS – 4/5 Development Methodology
Object Oriented Development Artifacts (e.g. UML diagrams) Validation
Few theoretically validated Mostly empirically validation (either data from industry or
student projects) Domain dominant
Data dominant (83%) Data dominant and control dominant (15%) Computational domain (2%)
13
RESULTS AND DISCUSSIONS – 5/5 Other aspects of software sizing
Critical reviews and improvements suggestions based on theoretical basis
How to calibrate Function Point Analysis (FPA) method Reliability of size measurement Convertibility between different size measures Tools for automated size measurement
14
CONCLUSIONS Software Product Size Measurement
Functional Size (86%) Physical Size (14%)
Only 19% are new/novel software sizing methods. Rest are either tailored or extended
Promising research areas: Tailoring the method for OO methodologies Tool development for automated measurement
Data dominant (83%), Data dominant and control dominant (15%), Computational domain (2%)
15
FUTURE WORK• A comprehensive Systematic Literature Review for in
depth analyses• Physical size measurement does not have similar level
of acceptancy. • Only 2% methods are for computational dominant
domain.
16
THANK YOU!!
Sohaib Shahid Bajwa ([email protected])
Cigdem Gencel ([email protected])
Pekka Abrahamsson ([email protected])
17