Abstract
This paper describes DARE, an architecture-based, decentralized framework that provides both self-configuration and self-healing properties to large and highly dynamic software systems. The design of DARE is based on the autonomic computing MAPE-K loop model and integrates architecture discovery mechanisms with recovery and adaptation connectors. This paper describes the design and detailed experimental validation of the DARE framework which show DARE's decentralized capability of dynamically adapting and recovering components.