Architecting Self-adaptive Software Systems

Abstract

A growing number of software systems operate in uncertain environments. They benefit from an ability to autonomously adapt to changes during runtime without suffering from a lowered quality of service. Several different architectural approaches to self-adaptive software exist with their sources of inspiration varying from psychology to mathematics. In this literature survey, we study and evaluate four types of approaches: architecture-based, control-based, learning-based and awareness-based approaches. Our aim is to clarify whether a unified, general approach to computational self-adaptivity is possible and what it could look like. We conclude that a general solution should combine aspects of all of the studied approaches.