Multi-Path QoS-Aware Service Composition

The goal of QoS-aware service composition is to generate optimal composite services that satisfy QoS requirements defined by clients. However, when compositions contain multiple execution paths, it is difficult to generate a solution that simultaneously optimizes all the execution paths involved in the composition at the same time, while meets global QoS constraints imposed by clients. Furthermore, finding an exact optimal solution by evaluating all possible combination, results in a combinatorial problem which is known to be NP-hard. This paper aims to tackle these major issues by proposing an innovative approach that suits for multi-paths compositions. In this approach, an optimization mechanism is proposed that computes the optimization by considering only the path that will be followed during the execution. To do so, a data mining method is adapted to predict, at runtime and just before the actual execution, the path that will be followed during the execution. By optimizing only the predicted path, it is expected to generate an optimal solution that delivering the best possible QoS performances, while guarantee meeting global QoS constraints. For solving the optimization problem, the problem is modelled as multi-dimensional multi-choice knapsack problem (MMKP). Then, heuristic algorithms are applied to solve it. By applying heuristic, it is expected to significantly reduce the computation efforts.