Research Of Data Model And Matching Algorithm On Semantic Based Publish/Subscribe System

With the rapid development of Internet, the network computing has come into the large-scale distributed computing age. Comparing to the traditional computing environment, the recent one is characterized by the large-scale, decentralization, dynamic and loose coupling. Since publish/subscribe systems have the advantage of fully decoupling in time, space and flow, they have become the infrastructure platform of the new distributed computing.The publish/subscribe systems have harvested a lot in content-based data model, matching algorithms and routing algorithms, but they are still short of support of semantic. The thesis does research on semantic-based data model and matching algorithms, the key technique of publish/subscribe systems, and develops a prototype of publish/subscribe system called JTang semantic publish/subscribe system, short for JTangPS, which strongly supports large-scale internet computing environment.At first, the thesis proposes a semantic-based data model of publish/subscribe system. The data model regulates the ability of expression. The data model of publish/subscribe system includes conceptual model, event model and subscription model. The conceptual model show the concepts of a domain and their relations, represented by OWL(Web Ontology language).The event model illuminates the event expression way, represented by RDF(Resource Description Framework) graph. The subscription model shows the expression way and filter condition of subscription, represented by RDF graph pattern. So the information expressed by data model could be exchanged between different applications, be understood and disposed by system. And more, we propose a new subscription language called JTangPS-SL.Base on the semantic data model, an efficient matching algorithm is required. The basic idea of matching algorithms is: decompounding the subscription graph pattern and event graph into a series of arcs which are the basic units for matching algorithms. Making full use of the overlap of subscription, merge lots of subscription into a large one. At the time of transforming a event, a reasoning mechanism is used to judge the validity of event and to produce extra information for matching with subscription. In the process of matching, we make use of the overlap of restriction to reduce the check of variable restriction and to make the process faster.At last, we develop a publish/subscribe system called JTangPS, which implemented the proposed data model and matching algorithms. The prototype is developed in Java, based on a exist overlay network and offer multi-way of accesing.