A Service Search Method Based On Referral Network

Service oriented computing is becoming a new distributed computing model. Web service is the effective mechanism of data and information integration on Web. At present, the network is moving in an open, dynamic environment, how to select a service becomes a very important problem under such environment. In web service system architecture, the main method of service discovery is to use UDDI. Although the centralized service searching method has played an important role at the beginning period of service composition, it is becoming the bottleneck of the system with the development of applications. In order to avoid the drawbacks of centralized service discovery, some solutions based on P2P network have been emerged. The referral network is a P2P system based on autonomous agents. This thesis studies service search related techniques based on referral network.Firstly, for solving the problems of service selection existing in the open, dynamic environment, this thesis proposes a Self-organization Service Referral Network Model ( SSRNM ) , then gives related policies and construction algorithms. Self-organization Service Referral Network Model could be constructed using these policies. The Self-organization Service Referral Network model provides the search space for service searching and composition.Secondly, a service research method based on SSRNM is provided which including Service Search Basic Algorithm and Service Search Augment Algorithm. Service Search Basic Algorithm can improve the success rate obviously by using referral. Service Search Augment Algorithm can solve network load balance trouble by introducing ant routing algorithm. In addition, the dynamic management mechanism of Self-organization Service Referral Network has been put forward.Finally, the experimental design and development have been carried out. The experimental results prove the SSRNM and the service search method are correct and feasible. And the experimental results show that the presented model can effectively improve searching efficiency.