Safe Runtime Validation of Behavioral Adaptations in Autonomic Software

Although runtime validation and verification are critical for ensuring reliability in autonomic software, research in these areas continues to lag behind other aspects of system development. Few researchers have tackled the problem of testing autonomic software at runtime, and the current state-of-the-art only addresses localized validation of selfadaptive changes. Such approaches fall short because they cannot reveal faults which may occur at different levels of the system. In this paper, we describe an approach that enables system-wide runtime testing of behavioral adaptations in autonomic software. Our approach applies a dependency-based test order strategy at runtime to facilitate integration and system-level regression testing in autonomic software. Since validation occurs on-line during system operations, we perform testing as part of a safe approach to adaptation. To investigate the feasibility of our approach, we apply it to an autonomic communication virtual machine.