Quality-aware analysis in product line engineering with the orthogonal variability model

Abstract

Software product line engineering (SPLE) is about producing a set of similar products in a certain domain. A variability model documents the variability amongst products in a product line. The specification of variability can be extended with quality information, such as measurable quality attributes (e.g., CPU and memory consumption) and constraints on these attributes (e.g., memory consumption should be in a range of values). However, the wrong use of constraints may cause anomalies in the specification which must be detected (e.g., the model could represent no products). Furthermore, based on such quality information it is possible to carry out quality-aware analyses, i.e., the product line engineer may want to verify whether it is possible to build a product that satisfies a desired quality. The challenge for quality-aware specification and analysis is three-fold. First, there should be a way to specify quality information in variability models. Second, it should be possible to detect anomalies in the variability specification associated with quality information. Third, there should be mechanisms to verify the variability model to extract useful information, such as the possibility to build a product that fulfils certain quality conditions (e.g., is there any product that requires less than 512MB of memory?). In this article, we present an approach for quality-aware analysis in software product lines using the orthogonal variability model (OVM) to represent variability. We propose to map variability represented in the OVM associated with quality information to a constraint satisfaction problem and to use an off-the-shelf constraint programming solver to automatically perform the verification task. To illustrate our approach, we use a product line in the automotive domain which is an example that was created in a national project by a leading car company. We have developed a prototype tool named FaMa-OVM, which works as a proof of concepts. We were able to identify void models, dead and false optional elements, and check whether the product line example satisfies quality conditions.

Supporting distributed product configuration by integrating heterogeneous variability modeling approaches

“…-we have implemented the Invar service configuration interface for the featureoriented FaMa tool suite [29], the OVM-oriented FaMa tool [43,44] tool suite. We chose these approaches as we have gained several years of experience of applying them in academic and industrial settings including large-scale product lines [29,36,14,38,45,39].…”

“…FaMa-OVM [44] is an instance of the FaMa framework [29] that uses the OVM notation [8]. The same analysis capabilities and extensions can be supported as for feature models (see Section 7.1.1).…”

Defining and Validating a Multimodel Approach for Product Architecture Derivation and Improvement

“…Several approaches for the quality evaluation and analysis of SPL architectures have been proposed over the last few years (e.g., [23], [28], [19], [31]). Among them, Kim et al [23] and Olumofin and Misic [28] propose two extensions of ATAM (i.e., EATAM [23] and HoPLAA [28]) with which to assess the quality of both product line and product architectures.…”

“…Although several methods for architecture derivation and improvement in SPL development have been proposed over the last few years (e.g., [23], [28], [19], [31], [6], [8], [29]), there is still a need for approaches that model the impact between architectural design decisions and quality attributes and use this information to enhance the quality attribute levels of product architectures. We have addressed this problem, in previous works [17] [18] [20], by proposing an approach with which to ensure the desired quality attribute levels for a product by applying architectural transformations to a product architecture derived from a product line architecture using a multimodel.…”

Validating a model-driven software architecture evaluation and improvement method: A family of experiments

“…Various studies concerned with the derivation (e.g., [14], [61]) and/or evaluation of software architectures from several points of view (e.g., [65], [37], [70], [66]) have been proposed in literature. After reviewing these studies, we have observed that: (a) There is a lack of systematic methods that model the impact between architectural design decisions and quality attributes to support the integrated derivation, evaluation and quality enhancement of software architectures.…”