Quality constraint driven local optimization for efficient service composition

Service composition can aggregate atomic Web services developed independently by their providers. However, with the increasing number of candidate services, the solutions based on global optimization and constraints cost much computation time, which make them inappropriate for real-time applications. In this paper, a heuristic service composition method, named LOEM-T (Local Optimization and Enumeration Method with Solution Tendency Estimation), is proposed. It aims to keep a small number of promising candidates for each task in service composition. As stringent constraints may lead to no feasible solutions, a notion of stringent degree, that can describe the influence of global constraints, is proposed. Then with the help of stringent degree, the candidates that are more excellent in the stringent constraints are prone to be preserved. The results of experimental evaluation shows that our approach is feasible and can achieve a near-optimal solution very efficiently.