Support for concurrent adaptation of multiple Web service compositions to maximize business metrics

Runtime adaptation of Web service compositions can often be done in several ways, so it is necessary to decide which adaptation approach to take. While many research projects studied runtime adaptation of Web service compositions or business processes, this paper presents our unique solutions that maximize business metrics, in cases when several Web service composition instances should be adapted at the same time. We specify all necessary information about possible adaptations and their business metrics as policies in our WS-Policy4MASC language and model the optimization problem in the powerful constraint programming language MiniZinc. Into our MiniZnMASC middleware we integrated new algorithms that determine how to adapt each Web service composition instance so the total business value is maximized, while satisfying all given constraints (e.g., about resource limitations). Experiments with the MiniZnMASC prototype showed that our solutions are feasible, functionally correct, business beneficial, with low performance overhead, and with linear scalability.