Dynamically adapting tuple replication for managing availability in a shared data space

With its decoupling of processes in space and time, the shared data space model has proven to be a well-suited solution for developing distributed component-based systems. However, as in many distributed applications, functional and extra-functional aspects are still interwoven in components. In this paper, we address how shared data spaces can support separation of concerns. In particular, we present a solution that allows developers to merely specify performance and availability requirements for data tuples, while the underlying middleware evaluates various distribution and replication policies in order to select the one that meets these requirements best. Moreover, in our approach, the middleware continuously monitors the behavior of application and system components, and switches to different policies if this would lead to better results. We describe our approach, along with the design of a prototype implementation and its quantitative evaluation.