A fault deployment and activation approach for testing self-recovery mechanisms

Abstract

The hypotheses on potential sources of error in an application can be specified in a fault model, which is useful for testing and for fault detection mechanisms. Based on a fault model, we developed custom mechanisms for providing self-recovery behavior in a component platform when third-party components behave inappropriately. In order to perform the tests for validating such mechanisms, it would be necessary to use a technique for fault injection so we could simulate faulty behavior. However such a technique may not be appropriate for a component-based approach. The behavior of systems tested with faults injected in the interface level (e.g., passing invalid parameters) would not represent actual application usage, thus significantly differing from cases where faults are injected in the component level (e.g. emulation of internal component errors). This paper presents our approach for testing, involving a general model for fault deployment and activation. Faulty components deployed at runtime represent faulty behaviors from the fault model. These faults are remotely activated through test probes that help testing the effectiveness of the platform's self-adaptive mechanisms.