A QoS registry for adaptive real-time service-oriented applications

Real-time service-oriented applications are characterized by Quality of Service (QoS) requirements that cannot be properly managed by using classical real-time systems methodologies. In dynamic environments the QoS management can effectively leverage adaptive techniques, that provide flexibility and do not require a complex offline analysis. In turn, such techniques make a massive use of on-line collected data regarding the application performance and the resource requirements. Moreover, a common issue for adaptive systems is the one of deciding the initial configuration of the application and/or the run-time environment at the time of service instantiation. In this paper, we propose a QoS registry for coping with these issues and supporting the configuration of proper scheduling parameters in real-time Service Oriented Architectures (SOAs). The registry permits to gather QoS data related to different functional behaviors of applications, to predict the future trend based on data already collected and to permanently store such data for an effective reuse at the time of future re-instantiations. We have also built an implementation of such registry, computed its overhead costs and performed some experiments for showing the effectiveness in auto-tuning resource allocations for providing QoS guarantees in a real-time SOA.