Chinese Journal of Computers   Full Text
Title	A Dependable and Adaptive Approach to Supporting Web Service Composition
Authors	GUO Hui-Peng HUAI Jin-Peng DENG Ting LI Yang
Address	(School of Computer Science and Engineering, Beihang University, Beijing 100083)
Year	2008
Issue	No.8(1434—1444)
Abstract &   Background	Abstract Service composition which integrates the functionalities of different services is a promising technique for developing applications especially for across multiple organizations. The dependability of Web services, however, is limited in some important ways for the distributed, dynamic and autonomous service domains. This paper addresses this problem by proposing dependable and adaptive approach for Service Composition. This paper initially transforms the composite services dependability maintain problem to a adaptive control problem, modeling the control process as a Markov decision process and therefore, it proposes an adaptive control system architecture, then designs and optimizes the maintenance strategy in accordance with control goals given in the setting of the theory of Markov Decision Process. It further gives Reinforcement Learning Based Adaptive Control Mechanism and corresponding algorithm to maintain the dependability of composite service. Finally this paper implements a prototype system to evaluate the proposed approach through comprehensive experiments and achieves improved results. Keywords service composition; dependability; adaptive control; reinforcement learning; Q learning algorithm Background This work is supported by the National Natural Science Funds for Distinguished Young Scholar under grant No.60525209; National Basic Research Program of China (973 Program) under grant No.2005CB321803 and the National High Technology Research and Development Program (863 Program) under grant Nos.2006AA01Z19A, 2007AA010301.The research scopes cover several fields of service computing, including service composition, service management, service engineering etc. The team has made important progress, amongst of which is the proposition of hierarchical service composition framework based on service overlay network (HOSS), which can be used to construct more general-purpose service overlay network through describing the relations among services using business protocols. Furthermore, algorithms are designed to support service selection for QoS-aware service composition. The work described in the paper is aimed at composing Web services to support dependable and adaptive business process. In the closed-loop control architecture for adaptive maintenance of composite services, maintain strategies are optimized in accordance with control goals given in the setting of the theory of Markov Decision Process. Q-Learning Based algorithm is designed for optimal control on the dependability of composite services. Thus the composite service can be maintained dynamically.