A data integration approach to automatically composing and optimizing web services

In this paper we show how data integration techniques can be used to automatically compose new web services from existing web services. A key challenge is to optimize the execution of the composed web services. We introduce a novel technique termed tuple-level filtering that optimizes the execution of the composed web services by reducing the number of web service requests. Moreover, we combine the tuple-level filtering algorithm with a technique that includes additional web service requests in the composition in order to improve filtering and further optimize the execution. Our initial experimental evaluation shows that our optimization techniques can reduce the execution time of the composed web services by up to two orders of magnitude. 1