Studying the Propagation of Failures in SOAs

Although Service Oriented Architectures (SOAs) are being increasingly used in business-critical scenarios, the applicability of Verification and Validation (V&V) is still very limited. The problem is that V&V activities have to be implemented at runtime to fit the characteristics of SOA. Recent proposals of runtime V&V techniques specific to SOA domain are far from being complete and a key issue lies in understanding how the "failures propagate" in a dynamic system and how to continuously verify its evolving elements. This paper introduces an approach to deal with the propagation of failures in a SOA environment. The proposed technique is based on three key steps: estimating the failure rate of the individual services, using fault injection to find the exposure of each service to failures from the invoked services, and estimating the impact of each service in the overall architecture. The overall approach is presented with a brief demonstration of its application.