Software Architecture Optimization Methods: A Systematic Literature Review

Reviewer: Hassan Reza

Software architecture has an important role in solving the problems of today's complex systems, especially the difficulty of manually searching and selecting the optimal solution from an exponentially growing design space. There is a gap in the existing work on software architecture optimization research. The authors of this paper present a comprehensive literature review of 188 papers in the area of software architecture optimization, using a systematic survey and screening process to include the most relevant papers in this area. Some contributions of this work include the screening guidelines used to rank and select the most relevant and significant works related to the optimization of software architecture using artificial intelligence and conventional methods, and the exclusion criteria used to eliminate candidate papers that focus on architecturally insignificant problems. The paper also introduces an evaluation technique that uses a well-designed taxonomy scheme to classify the existing optimization works into three major categories: problem (how the problem is represented or formulated), solution (the solution techniques that were used), and validation (how the approach was evaluated or tested). The authors identify gaps in the existing research and propose new directions to improve the state of research in the architectural optimization area. The significance of this survey work builds on the systematic literature review proposed originally by Kitchenham [1]. The goal of the review protocol is to provide proper answers to the fundamental survey questions formulated by the authors. To answer these questions, the literature review process employed three phases: planning (or setup), conducting (or data extraction), and reporting (identifying the gap and proposing new research directions). The planning phase (setup task) includes the definition of the review protocol, the selection of search engines, the development of a keyword list with definitions, a keyword-based collection of published architecture optimization papers, and filtering mechanisms to include or exclude papers. The content of the selected papers was then analyzed to create the taxonomy for classifying existing approaches to architectural optimization. An iterative coding process was then used to update the initial taxonomy. The output of the planning phase is used as an input to the conducting phase to extract and check data. The main objective is to collect and validate significant works in the area of architectural optimization. The reporting phase involves a cross-analysis of the survey results to identify holes in the existing work and to propose new research directions for future work in this particular area. The authors conducted an extensive literature review to systematically select significant work in the area of software architecture optimization. The taxonomy and process used to identify and document this body of work is very rigorous. For instance, each major category (problem, solution, and validation) has been further classified by subcategories (for example, domain, quality evaluation, and so on) to incorporate various aspects of optimization based on the software architecture of a system. Thirteen tables were devised to properly document all of these aspects of optimization approaches using the taxonomy framework. This paper is well written and well organized. I recommend it to graduate students who are eager to pursue research in the areas of software architecture, architectural design tools, architectural tradeoff analysis, or quality-oriented software architecture. Online Computing Reviews Service