Cross organizational service based workflows - solution strategies for quality of service optimization

By the application of the Service-oriented Architecture (SOA) paradigm on business processes, workflows can be decomposed into basic activities that can be realized by reusable services offering a specific business functionality. In order to compose cross-organizational service-based workflows, services can be sourced from internal as well as from external providers. On a large service market, services are offered with varying Quality of Service (QoS) levels and several pricing models. Providing a high level of QoS concerning composite services or service-based workflows is of high importance for an intermediary, acting as a service orchestrator, selling composed service-based workflows to his consumers. Besides efficient workload predictions, fast and efficient solution strategies for QoS and cost optimization are mandatory for the intermediary in order to stay competitive and to enable fast reaction strategies on varying demands of workflow execution requests. This thesis provides several contributions to the QoS optimization of servicebased workflows from the intermediary’s point of view. The main contribution is the development and the evaluation of efficient resource planning heuristics, facilitating the fast computation of invocation plans out of services with limited execution capacities, offered by a specific pricing model. Thus, a resource planning optimization model, solved by mathematical optimization with an exact solution as well as by the application of the developed heuristics, is introduced. Additional contributions address several challenges in the field of QoS optimization of service-based workflows. As a foundation, a classification of pricing models for services is developed and the impact of several pricing models on the service selection process for service-based workflows is presented. Several solution strategies for the QoS optimization are developed supporting the intermediary in the worstand average-case performance analysis of servicebased workflows. In an average-case analysis, key findings of queuing theory are adapted to the concept of service-based workflows and several optimization approaches are developed. These support the intermediary by the optimization of the service utilization incorporating constraints such as the overall response time. Furthermore, major concepts of network calculus are identified and adapted to the concept of service-based workflows. Consequently, optimization approaches are developed facilitating the optimization of QoS parameters such as the delay or the throughput in the worst-case. Finally, an architectural extension for generic QoS management systems for service-based workflows is proposed, facilitating the realization and implementation of the developed solution strategies for the resource planning of service-based workflows.