Improving service selection in component-based architectures with optimal stopping

This paper describes how optimal stopping strategies are applicable to certain situations of service selection in component-based and service-oriented software architectures. It shows how optimal stopping can be used at runtime to stop with 0(n) complexity a dynamic service selection process and then yield 1/e probability for the best choice. It provides experimental confirmation in a simulation with currency conversion Web services. Areas of further applicability include grid computing, distributed multimedia and mobile computing.