Asad Abbas ¹, Isma Farah Siddiqui ¹, Scott Uk-Jin Lee ¹

Affiliations
1 Department of Computer Science and Engineering, Hanyang University, South Korea

Abstract

Software Product Line (SPL) is process for developing families of software with reusability of features categorized as common and variable features. Feature Model (FM) is developed to manage these features. Common features are easy to manage, however variable features are hard to manage because of complex relations and constraints between features. Optimization is required to manage the variabilities for best selection of features and product configurations. To this end, different Multi-Objective Evolutionary Algorithms have been proposed to get the optimal solutions of feature model. In this paper we have compared among three main optimization algorithms i.e. IBEA, NSGA-II and MOCell. Our comparison is based on previous research correctness solutions for product’s configuration with five objective functions on different feature models from SPLOT and LVAT repositories. The goal of this comparison is to find the current research prospective and challenges of multi-objective optimization in FM.

Keywords

Feature Model, Optimization of Feature Model, Software Product Line.

Full Text

   

Asad Abbas ¹, Zhiqiang Wu ¹, Isma Farah Siddiqui ¹, Scott Uk-Jin Lee ¹

Affiliations
1 Department of Computer Science and Engineering, Hanyang University, KP

Abstract

In Software Product Line (SPL), feature model is highly recommended to manage the commonalities and variability of features under resource constraints of mandatory, optional and alternative. Features with mandatory constraints and high in dependency with other features are identified as crosscutting concerns; reduce the reusability of resources. It is important to find and modularize these concerns at modeling level. With this practice, these concerns do not effect if deletion or addition is required from entire system. In this paper we have applied Union-find algorithm to find crosscutting concerns in feature model. We evaluated our approach by applying on an automobile feature model with various dependencies between features, and found required crosscutting concerns. By this approach, identification of crosscutting concerns and their modularization made easier. Further, we have also applied genetic algorithm to get optimized feature selection under cost constraint with high performance. In SPL, as crosscutting concerns are mandatory features with fix cost and performance, optimization on feature model is necessary under consideration of crosscutting concerns. Our approach found all possible products according to crosscutting concerns, cost and performance at modeling level of an automobile feature model. At last, we found all products from minimum to maximum cost with respect to least maximum performance by using GA optimization technique.

Keywords

Feature Model, Genetic Algorithm, Optimization, Software Product Line, Union-find Algorithm

Full Text

   

Isma Farah Siddiqui ¹, Asad Abbas ¹, Abdul Rahim Mohamed Ariffin ¹, Scott Uk-Jin Lee ¹

Affiliations
1 Department of Computer Science and Engineering,Hanyang University ERICA, Ansan, South Korea, KR

Abstract

In the infrastructure of Information and Communication Technology (ICT), the design and architecture of software deployed over equipment plays a vital role in system’s concurrent efficiency. The high performance computing nexus with fine-grain parallel processing environment calls for optimized and effective multithreading strategies for ICT’s software implementations. This paper briefly identifies different state-of-art multithreading APIs used in multi-core coherent processors. The paper investigates various APIs based on main categorization of operating system level and lightweight thread level. Operating system level multithreading relies on operating system kernel architecture as numerous threads or tasks achieve concurrency provided by the underlying hardware, thus considered effective for high core programming. On the other hand, various application level lightweight thread models are been offers with lighter mechanism for high parallelism and massive concurrency. Lightweight models are optimized to combine low-latency thread and task scheduling with optimized functionality for data-movement. This comparative study aims to demonstrate major working models and principal differences among well-established libraries in each category. This paper projects an early investigation for the identification of most efficient thread library for ICT equipment.

Keywords

ICT Software, Parallelism, Multithreading Models.

Full Text