Applying particle swarm optimization to software performance prediction an introduction to the approach

Abstract

Component-Based System (CBS) is an approach to build applications from deployed components. It provides efficiency, reliability, maintainability. The challenge of interpreting the results of performance analysis and generate alternative design to build component system is quite critical in the software performance domain. Although, many approaches have been proposed and were successfully applied to predict software performance, still span of design space hinder the selection of the appropriate design alternative. Meta-heuristics such as Genetic Algorithms (GA) methods have proven its usefulness to solve the problem even with multi-degree of freedom. But, in recent investigations Particle Swarm Optimization (PSO), an alternative search technique, often outperformed GA when applied to various problems. In this paper we describe performance prediction approach based on PSO for component-Based system development. The proposed approach aids developers to effectively trades-off between architectural designs alternatives. Boundary search technique and PSO are used to provoke more efficient results. To the best of our knowledge we are the first who employ PSO in software performance prediction. Outlines of our approach are presented and a case study applied using GA is described to be used by our approach for validation. This paper has concluded that, PSO technique can be used to effectively generate alternatives in spanned design space and facilitate the design decision during the development process.