Optimizing Knowledge-based Software Architecture Designs Using Genetic Algorithms

摘要

Software architectural design is one of the biggest challenges in the research field of modern software engineering. The knowledge produced and consumed during the development of architectures is capable of serving architects in the lifecycle phases. Thus, architectural knowledge (AK) plays a vital role in the design, reuse, and evolution of software development. The unavailability of AK precludes organizations from growing their architectural capabilities. Ontology-based representation of rationale and architectural concepts groups all of the metadata concerning the software architecture and guides the architect in developing the most appropriate architectures, where the interactions between quality attributes and tactics are analyzed for the quality-driven architectural design process. Identifying specific software solutions that contribute to achieving the business requirements increases the efficiency of the software development process. However, the quality properties are mostly conflicting, adding to the complexity of the problem. A brute-force approach for finding the optimal solution would result in an exponentially greater effort for the rising number of architectural solutions. In this article, a framework is proposed for optimizing the architectures using heuristics. Genetic algorithm (GA) is a promising heuristic approach to converge on a subset of acceptable Pareto optimal solutions to the multi-objective problems in large spaces. Use of AK and GAs reduces the development time, thereby enhancing the quality of the architectural design process.