Software engineering processes for self-adaptive systems

Abstract

In this paper, we discuss how for self-adaptive systems some activities that traditionally occur at development-time are moved to run-time. Responsibilities for these activities shift from software engineers to the system itself, causing the traditional boundary between development-time and run-time to blur. As a consequence, we argue how the traditional software engineering process needs to be reconceptualized to distinguish both development-time and run-time activities, and to support designers in taking decisions on how to properly engineer such systems. Furthermore, we identify a number of challenges related to this required reconceptualization, and we propose initial ideas based on process modeling. We use the Software and Systems Process Engineering Meta-Model (SPEM) to specify which activities are meant to be performed off-line and on-line, and also the dependencies between them. The proposed models should capture information about the costs and benefits of shifting activities to run-time, since such models should support software engineers in their decisions when they are engineering self-adaptive systems.