Towards architecture-based self-healing systems

Our approach to creating self-healing systems is based on software architecture, where repairs are done at the level of a software system's components and connectors. In our approach, event-based software architectures are targeted because they offer significant benefits for run-time adaptation. Before an automated planning agent can decide how to repair a self-healing system, a significant infrastructure must be in place to support making the planned repair. Specifically, the self-healing system must be built using a framework that allows for run-time adaptation, there must be a language in which to express the repair plan, and there must be a reconfiguration agent that can execute the repair plan once it is created. In this paper, we present tools and methods that implement these infrastructure elements in the context of an overall architecture-based vision for building self-healing systems. The paper concludes with a gap analysis of our current infrastructure vs. the overall vision, and our plans for fulfilling that vision.