Search‐based many‐to‐one component substitution

Abstract

SUMMARYIn this paper, we present a search-based automatic many-to-one component substitution mechanism. When a component is removed from an assembly to overcome component obsolescence, failure or unavailability, most existing systems perform component-tocomponent (one-to-one) substitution. Thus, they only handle situations where a specific candidate component is available. As this is not the most frequent case, it would be more flexible to allow a single component to be replaced by a whole component assembly (many-to-one component substitution). We propose such an automatic substitution mechanism, which does not require the possible changes to be anticipated and which preserves the quality of the assembly. This mechanism requires components to be enhanced with ports, which provide synthetic information on components' assembling capabilities. Such port-enhanced components then constitute input data for a searchbased mechanism that looks for possible assemblies using various heuristics to tame complexity.key words: component substitution, component assembly evolution, search-based building process, many-to-one component replacement, heuristics, dead components Introduction Nowadays, software systems have to meet the needs of long life, autonomous and ubiquitous applications and must therefore be flexible, dynamic, and adaptable like never before. * Correspondence to: LGI2P, Ecole des Mines d'Alès, Parc scientifique Georges Besse, F30035 Nîmes cedex, France † E-mail: Nicolas.Desnos@ema.fr, huchard@lirmm.fr, tremblay.guy@uqam.ca, Christelle.Urtado@ema.fr, Sylvain.Vauttier@ema.fr Component-based software engineering (Cbse) [1] is a good solution to optimize software reuse and dynamic evolution while guaranteeing the quality of the software. Typically, a component is seen as a black box which provides and requires services through its interfaces. An architecture is built to fulfill a set of functional objectives (its functional requirements) while enforcing a set of properties (its non-functional requirements) and is described as a static interconnection of software component classes. A component assembly is a runtime instantiation of an architecture composed of linked component instances.In long life applications or evolving environments, component substitution is a necessary mechanism for software evolution: it is a response to such events as component obsolescence, failure or unavailability. Anticipating valid component substitutions while designing some software is not always possible as the various contexts in which that software may run are not known in advance. Repairing a component assembly after a component has been removed while still preserving its whole set of functionalities is thus difficult. When a component is removed from an assembly, most existing approaches perform component-to-component (oneto-one) substitution [2,3,4,5]. However, these approaches rely on the fact that an appropriate component, candidate for substitution, is available. This situation can hardly happen because it is difficult to fin...

An optimization‐based tool to support the cost‐effective production of software architecture documentation

“…Optimization techniques have been widely used in several fields of Software Engineering, including Requirements Engineering, Software Design and Development, Software Maintenance and Evolution, or Project Planning, among others . In the context of Requirements Engineering, Bagnall et al stated the NRP, which has several similarities with our SAD optimization problem.…”

“…When one or more components become obsolete, they must be replaced with one or more new components (e.g., COTS, Commercial‐Off‐The‐Shelf) that have similar characteristics and functionality. Desnos et al have studied this combinatorial problem and evaluated the performance of several optimization techniques using synthetic data. Another CBSE problem that is suitable for SBSE techniques is the selection of components for the next version of a software system, which is a similar problem to NRP.…”

Search-based software engineering

“…Baker et al [2006] presented results on greedy optimization and SA for component selection, while Yang et al [2006] proposed an approach for the software integration problem by using GAs to reduce risk. Classical OR techniques have also been applied to component selection problems: Desnos et al [2008] combined backtracking and branch-and-bound techniques for automatic component substitution problem to optimize software reuse and evolution. Other authors have considered the component selection problem as a selection optimization problem.…”

Automated architectural component classification using concept lattices

“…Finally, at runtime, dynamic autonomous software might need to re-assemble some of its parts to react to component failure or unavailability. During previous work on automatic component assembly and on dynamic component assembly evolution [1], we recognized that an efficiently indexed component directory was a central issue for component reuse.…”