Distributed Service Composition in Internet of Services

Abstract

To meet coarse-grained user requirements, many service composition approaches have been put forward in the past 20 years. The service composition constructs a service workflow by aggregating services to meet users’ functional and non-functional requirements. In the era of the Internet of Services (IoS), service solutions are constituted by services deployed on distributed platforms and owned by different providers. For reasons such as commercial confidentiality, different providers or platforms (called "IoS nodes") generally do not share complete information of services to other nodes. Thus, how to efficiently construct optimal service composite solutions for complex user requirements under this distributed service environment is a great challenge. In this paper, we propose a three-stage distributed service composition method with coordination protocols and local algorithms. In the first stage, when an IoS node receives a requirement (the initial node), it uses a centralized service composition algorithm and broadcasts sub-requirements it can’t meet to all other nodes through a requirement broadcasting protocol. In the second stage, the initial node uses a node selection algorithm to assign a node to provide the service for each sub-requirement. In the third stage, the nodes coordinate their services with other nodes through a constraint coordination protocol based on the upper confidence bound algorithm and the initial node obtains a composite service solution. The experimental results demonstrate that our method outperforms other distributed service composition methods in terms of efficiency and solution quality.