A Case Study on Architecture-based Resilience Evaluation for Self-Adaptive Systems

Abstract

One of the most promising approaches to face the increasing complexity of software systems is the use of self-adaptation, in order to enable software systems to deal with changes themselves, autonomously. It is presented as one of the means by which it is possible to provide systems that are scalable, support dynamic modifications and rigorous analysis, capable to respond to resource variability or user needs modifications, still, being flexible and robust. Normally, by design, the methods for self-adaptation are at the system’s source code or network level, but recently, architecture-based methods have been widely considered as more promising approaches. One of the main barriers for greater implementation of architecture-based self-adaptation is the lack of evidence of the advantages and compensation of applying it in systems with built-in adaptation mechanisms. A recent proposal to cope with this challenge uses an architecture-based approach which evaluates alternative adaptation mechanisms of a self-adaptive system by comparison, based on the identification of representative system and environmental conditions which may have a relevant impact on system resilience. The present work has one major contribution: evaluate if the application of architecture-based self-adaptation can improve the resilience of an already adaptive system. The effectiveness of the above-mentioned approach is demonstrated by using Rainbow, an architecture-based platform for self-adaptation, and DCAS, an industrial software-intensive system used to monitor and manage highly populated networks of devices in renewable energy production plants. The experimental evaluation showed that the application of architecture-based self-adaptation improved the resilience of the tested system. The overall runtime quality of the self-adaptive system can be greatly improved with acceptable costs.