The availability and reliability of Service-oriented architectures (SOA) depends on two factors: On the one hand, the availability and reliability of the services that provide a certain business functionality and on the other hand the services that make up the underlying SOA platform. For platforms that are supposed to form the core of mission-critical service-oriented applications, this implicates the need for mechanisms that can regulate the reliability- and availabilitylevels of the core services in changing conditions. In this paper, we discuss open questions about what kind of monitoring functionalities and service replication mechanisms should be integrated in SOA infrastructures. Therefore, the integration of concepts from peer-to-peer (P2P) computing is proposed: We present a self-organization extension that can improve the availability of the core services of SOA infrastructures, and we provide an experiment-based evaluation, showing some of the benefits that this extension can have in a critical scenario. The concepts are prototypically implemented as extensions of Apache Tuscany, which is a realization of the Service Component Architecture (SCA) standard.
