An Adaptive Service Selection Approach to Service Composition

In service computing, the behavior of a service may evolve. When an organization develops a service-oriented application in which certain services are provided by external partners, the organization should address the problem of uninformed behavior evolution of external services. This paper proposes an adaptive framework that bars problematic external services to be used in the service-oriented application of an organization. We use dynamic WSDL information in public service registries to approximate a snapshot of a network of services, and apply link analysis on the snapshot to identify services that are popularly used by different service consumers at the moment. As such, service composition can be strategically formed using the highly referenced services. We evaluate our proposal through a simulation study. The results show that, in terms of the number of failures experienced by service consumers, our proposal significantly outperforms the random approach in selecting reliable services to form service compositions.