A Genetic Algorithm for Automated Service Binding

AbstractBinding concrete web services to the tasks involved in an orchestration model is an important step in dynamic web service composition. In our previous work on automated service binding, we have described a method of combining a powerful technique for computing the QoS of a composite service with our own approach of expressing preferences related to the trade-offs between the various QoS parameters of a composite service. In this paper, we introduce a genetic algorithm that finds the best mapping of concrete services to the tasks involved in the composition. The algorithm uses the method presented in our previous work in order to estimate the fitness of the mappings that make up a population of candidate solutions. We describe how the challenges posed by the use of this method have influenced the design of our algorithm and report the experimental results showing the effectiveness of our approach