Quality Attributes Assessment for Feature-Based Product Configuration in Software
Product Line
Guoheng Zhang, Huilin Ye, and Yuqing Lin
School of Electrical Engineering and Computer Science,
University of Newcastle
Callaghan 2308, NSW, Australia
Guoheng.Zhang@studentmail.newcastle.edu.au
{Huilin.Ye, Yuqing.Lin}@newcastle.edu.au


Abstract— Product configuration based on a feature model in
software product lines is the process of selecting the desired
features based on customers’ requirements. In most cases,
application engineers focus on the functionalities of the target
product during product configuration process whereas the
quality attributes are handled until the final product is
produced. However, it is costly to fix the problem if the quality
attributes have not been considered in the product
configuration stage. The key issue of assessing a quality
attribute of a product configuration is to measure the impact
on a quality attribute made by the set of functional variable
features selected in a configuration. Current existing
approaches have several limitations, such as no quantitative
measurements provided or requiring existing valid products
and heavy human effort for the assessment. To overcome
theses limitations, we propose an Analytic Hierarchical Process
(AHP) based approach to estimate the relative importance of
each functional variable feature on a quality attribute. Based
on the relative importance value of each functional variable
feature on a quality attribute, the level of quality attributes of a
product configuration in software product lines can be
assessed. An illustrative example based on the Computer Aided
Dispatch (CAD) software product line is presented to
demonstrate how the proposed approach works.
Keywords-quality attribtues assessment, product
configuration, Analytic Hierarchical Process (AHP).
I. INTRODUCTION
A software product line (SPL) enables to create a number
of similar products by selecting and composing the reusable
software artifacts [4]. The key aspect of developing a
software product line is to capture the commonalities and
variabilities of software product line members. The
commonalities and variabilities are identified and modeled
during domain analysis as features in a feature model [5].
Product derivation is the process of developing products
from a software product line. Product derivation starts with
product configuration in a feature model where the desired
features are selected from a feature model based on
customers’ requirements. Based on the selected features, the
reusable code units are selected and composed into the final
products. In most cases, application engineers and end-users
pay attention to functional variabilities which satisfy the
business needs of the system during product configuration in
a feature model. Non-functional requirements or quality
attributes of the products, such as performance, security,
usability and development cost, are usually handled until the
final product is produced and tested in the system testing
phase [1]. Different members of the software product line
may require different levels of quality attributes. For
example, one product may require a very high security
whereas in another product security is not that important. If it
is found that the quality attributes of the product fail to meet
the customers’ requirements in a later product development
stage, it is costly to fix the problems. Therefore, the quality
attributes of a target product should be assessed as early as
possible in the product development process. As the product
configuration in a feature model is the first stage of product
development, quality attributes should be assessed in this
stage.
Several approaches [2] [4] [5] [6] have been proposed to
address the issue of quality attributes assessment in software
product lines. The quality attributes are classified into
optional quality attributes, quality attribute levels, or impact
of functional variable features on quality attributes [4] [5]
[6]. The optional quality attributes and quality attribute
levels can be simply treated as features and selected directly
in a product configuration [4] [5]. The impact of functional
variable features on quality attributes is complex and
difficult to identify and measure, because one variable
feature may affect several quality attributes while one quality
attribute may be affected by several variable features. It is
necessary to identify and measure the different impact of
functional variable features on a quality attribute, because it
plays a key role in assessing the quality attributes for a
product configuration.
The impact of functional variable features on quality
attributes can be measured using either qualitative analysis
methods or quantitative analysis methods. In the qualitative
analysis methods [2] [8], the impact of functional variable
features on quality attributes are measured by some
qualitative labels, such as “positive”, “negative” or
“unknown”. The quality attribute level for a product
configuration is estimated by combing a set of qualitative
impacts. Thus, the qualitative analysis methods only support
a rough estimation. To provide more precise estimation, the
approaches in [3] [4] [5] [7] proposed quantitative analysis
methods. The impact of functional variable features is
mainly measured by comparing a set of valid products which
are difficult to obtain. The approach in [9] covers this
limitation and uses Bayesian Belief Network (BBN) to
represent the quantitative impact of a configuration on a
2010 Asia Pacific Software Engineering Conference
1530-1362/10 $26.00 © 2010 IEEE
DOI 10.1109/APSEC.2010.25
137