DISSERTATION QoS - Aware Composition of Adaptive Service - Oriented Systems

Service-Oriented Computing (SOC) increasingly gains momentum in academia and industry as a means to develop adaptive distributed software applications in a loosely coupled way. Software services, as the main entities in SOC, have some distinct properties such as platformindependence or a uniform interface description enabling an easier integration and use within and across organizational boundaries. One of the main assets of service-orientation is composability to develop higher-level services, so-called composite services, by re-using well-known functionality provided by other services in a low-cost and rapid development process. However, in distributed environments, the use of services without any quality guarantees from the service providers can negatively affect a composite service by raising intermittent failures or having a slow performance of one of the services. One of the main problems is the lack of an integrated Quality of Service (QoS) model combined with an automated monitoring technique. The availability of accurate and up-to-date QoS information enables a QoS-aware composition and optimization of composite services by automatically selecting well-performing services and dynamically replace services that reduce the performance or lead to failures in a composition. However, existing QoS-aware composition approaches mainly focus on the optimization aspect to find the best composition in terms of QoS. Therefore, QoS should be seamlessly integrated into multiple layers of the SOC stack, such as choreography, orchestration and execution. This enables an end-to-end view on QoS and allows a better integration and optimization throughout the application lifecycle to achieve the vision of adaptive service-oriented systems. This thesis contributes a set of methods and tools to address these issues. Firstly, it proposes an extensible multi-layer QoS model for services and an automated QoS monitoring approach. Secondly, it describes the integration of Service Level Agreements (SLAs) into choreographies and proposes an automated mapping to orchestrations annotated with QoS policies to enable SLA enforcement. Thirdly, it addresses a set of issues related to the overall development lifecycle of QoS-aware service composition. Specifically, a domain-specific language, called VCL, is introduced to enable the specification of QoS-aware composite services with a focus on hard and soft constraints in form of constraint hierarchies. Based on VCL, a set of methods and algorithms are presented to generate an executable composite service that is optimized with regard to the QoS constraints specified by the user. To this end, this thesis also introduces a novel Web service runtime environment, called VRESCO, which implements a number of important SOC concepts (such as dynamic binding or invocation) that are foundational for the presented QoS-aware service composition approach.