Redundancy can improve the availability of components in service-oriented systems. However, predicting and quantifying the effects of different redundancy strategies can be a complex task. In our work, we have taken an architecture based approach to the modeling, predicting and monitoring of properties in distributed software systems. This paper proposes redundancy strategies for serviceoriented systems and models services with their associated protocols. We derive formal models from these high-level descriptions that are embedded in our fault-tolerance testing framework. We describe the general framework of our approach, develop two service redundancy strategies and report about the preliminary evaluation results in measuring performance and availability of such services. While the assumptions for the chosen case study are limiting, our evaluation is promising and encourages the extension of our testing framework to cater for more complex, hybrid, fault-tolerance strategies and archi...
Nicholas R. May, Heinz W. Schmidt, Ian E. Thomas