Publish / Subscribe System Design And Implementation Of The Data Model

With the rapid development of network and communication technology, a great deal of data and information has been generated, which is distributed in different system. The high-efficiency, flexibility and on-demand of information distribution is needed for effective operation of the system, distributed technology can used to solve this problem. At present, the distributed systems show a trend of increasing scale and more and more complex structure, and have the characteristics of Decentralized control, dynamic, loosely coupled. The publish/subscribe paradigm can make the information Producers and consumers fully decoupled in time, space and flow, so it is well satisfied the demand of information distribution in the distributed systems.The publish/subscribe systems have achieved a lot in content-based data model, matching algorithms and routing algorithms, however, they are still short of expression ability. The thesis focus on the key technique of publish/subscribe systems which are based on XML, takes advantages of the low memory in XML stream parsing and the rich expression, designs the event model and subscription model, based on this, designs the matching algorithm, finally, constructs a prototype of publish/subscribe system, which provides infrastructure for information distribution of distributed systems.Firstly, the thesis constructs a data model of the publish/subscribe systems based XML, the data model decides the expression ability of the system, the data model includes event model and subscription model. The event model decides how to express the event, the event is described as a XML stream with index which is used to get the postfix nodes information in advance. The subscription model decides the expression of subscription and filter condition, the subscription is described as a Nondeterministic Finite Automaton corresponding to the XPath query expression, the NFA can achieve efficient filtering.Based on the data model described above, fast and efficient matching algorithm is needed. The core idea of this algorithm is to use the events generated by the XML parser to drive the transitions in the NFA to the accepting state, to get the subscriber corresponding to the accepting state. All of the queries are combined into a single, nondeterministic finite automaton in order to allow the sharing of processing for query paths with common prefixes and improve the time efficiency and space efficiency. In the process, the index is used to reduce the parsing redundancy and achieve fast matching.Finally, a prototype of publish/subscribe system is constructed which implement the data model and matching algorithm. The system is a part of the research of the key technology of the XXX and information distribution, which is a sub project of the key pre-research of the eleventh five-year plan, the system implement the key technology of the publish/subscribe systems.