An Adaptive Quality Level Selection Method for Efficient QoS-Aware Service Composition

In the task-oriented service computing framework where services are composed together to accomplish a task goal of a user, appropriate component services need to be dynamically selected and bound to the task. When there are many candidate services that provide similar functionality, it is essential to consider quality of services (QoS) such as response time, cost, availability, and reliability to decide which component services to use. Finding a service composite that meets the optimal quality is a well-known NP-hard problem because the time complexity for the global optimization increases exponentially as the number of services or the number of QoS attributes increases. Although there is a heuristic approach that shows a reasonable response time with a certain level of service quality, it often fails when the global QoS constraints become tight. In this paper, we propose an adaptive way of dividing quality levels where candidate services are sampled and their QoS values are evaluated. The range of a quality level is dynamically decided based on the distribution of candidate component services on each QoS attribute, and the tightness of the constraint requirement within a task. Evaluation results show that the proposed approach can successfully reduce the failure rate of service composition while keeping the computation time reasonably low and ensuring the QoS optimality of composite services.