Optimized decentralization of composite web services

Composite services are usually specified by means of orchestration models that capture control and data-flow relations between activities. Concrete services are then assigned to each activity based on various criteria. In mainstream service orchestration platforms, the orchestration model is executed by a centralized orchestrator through which all interactions are channeled. This architecture is not optimal in terms of communication overhead and has the usual problems of a single point of failure. In previous work, we proposed a method for executing service orchestrations in a decentralized manner while fulfilling collocation and separation constraints. However, this and similar methods for decentralized orchestration do not seek to optimize the communication overhead between services participating in the orchestration. This paper presents a method for optimizing the selection of services assigned to activities in a service orchestration in terms of QoS properties and communication overhead. The method takes into account the communication cost between pairs of services, the amount of data that these services need to exchange in the orchestration, and the collocation and separation constraints imposed by the service providers.