The provision of dependable computer systems by deploying diverse, redundant components
in order to mask or provide recovery from component failures has mostly been restricted to
systems with very high criticality. In this paper we present an architecture and prototype implementation of an approach to providing such redundancy at low cost in service-based infrastructures. In particular we consider services that are supplied by composing a number of component
services and consider how service discovery, automatic monitoring and failure detection have the
potential to create composed services that are more dependable than might be possible using a
straightforward approach. The work is still in its early stages and so far no evaluation of the
approach has been carried out.