Research On Key Matching Technologies Of Content-based Publish Subscribe System

With the rapid development of modern technology, due to its tight coupling, traditional di stributed systems fail to meet contemporary society’s needs for flexible, loose-coupling and dynami c communications models, while the optimum architecture to solve these needs is publish/subscribe system, which is generally divided into three categories:channel-based, theme-based and content b ased. Although the former two systems are simple to implement, they are confined to weak expres sion ability. The content-based publish/subscribe system has expressive and flexible subscription Ian guage and conducts fine-grained information retrieval, thereby becoming a heated research topic.The research background of this paper is based on the 2012 annual National Natural Scie nce Fund Project (61272544), "Research on the key technology of COBSATRR CBPS, content bas ed on matching project," as a reference to do related theory research to the matching problem in t he publish/subscribe system.Starting from the most common system, through certain theory learning about the two kin ds of models --- publishment and subscription, to understand the event notification service. And ex plain in detail of several typical models and publish subscribe model in the publish/subscribe syste m, then introduces some relevant terms in CBPS on this basis.Subscription interval tree differentiating algorithm design based on the golden section met hod, namely IDAGSM design, is proposed. The research background is adopted as the theoretical basis to improve the existing tree interval differentiating method on the COBASTREE. By compari son of experiments, this algorithm is validated to be capable of balancing and maintaining the cost, and improving the matching speed at the same time, avoiding the overload.A new type of matching tree structure-QuTREE tree, which is an ordered and balanced q uadtree, is proposed. It transfers the matching problem of the published content and the predicate i n subscription filter into a query problem of a data point stabbing on a region. And the common dichotomy is used to differentiate the region. Through the experiment comparing with the tradition al tree matching, experimental results show that QuTREE tree can effectively reduce the depth of the subscription interval tree, improving the efficiency of matching and reducing the forwarding co St.