Towards Change Propagating Test Models in Autonomic and Adaptive Systems

Abstract

Autonomic and adaptive computing systems can add, remove, and replace their own components in response to a changing environment. Self-adaptation facilitates the performance of automated maintenance and configuration tasks, but makes it possible for faults to be introduced into the software at runtime. To address this issue, researchers have developed approaches for integrating runtime testing into autonomic and adaptive software systems. An important aspect of runtime testing approaches for autonomic software is the provision of a framework for regression testing, which determines whether modifications have introduced faults into previously tested components. However, after adaptation occurs in autonomic software, a predefined test set may no longer be applicable due to changes in the program structure. Investigating techniques for dynamically updating regression tests after adaptation is therefore necessary to ensure such approaches can be applied in practice. In this paper we describe a model-driven approach that maps structural adaptations in autonomic software, to updates for its runtime test model. We provide a workflow and meta-model to support the approach, referred to as Test Information Propagation (TIP). To demonstrate TIP, we have developed a prototype that simulates a reductive change to an autonomic, service-oriented healthcare application. Conducting the simulation has provided us with much insight into this highly challenging research problem.