Techniques to Produce Optimal Web Service Compositions

As Web Services proliferate, it becomes more difficult to find a service that can perform a given task, and a coordination of several services may be required. We present two algorithms to identify orderings of Web Service compositions. These algorithms follow different strategies to prune the space of possibilities while minimizing the evaluation cost. The first one, DP-BF, combines a best first strategy with a dynamic-programming technique and produces good Web Service compositions by exploring a small portion of the search space. The second one, PT-SAM, adapts a petri-net unfolding algorithm and tries to find a desired marking from an initial state. We conducted an experimental study to evaluate the behavior of DP-BF and PT-SAM comparedto SAM and to the exhaustive solution. Our experimentsshow that the quality of the compositions identified by ouralgorithms are close to the optimal solution produced by theexhaustive algorithm, while the optimization time is close to the time required by SAM to identify a solution.