Research And Implementation Of System Optimazation And Parallel Query Algorithms For Homogeneous Symmetric Publish/Subscribe System

Publish/Subscribe system is a kind of a distributed middleware providing a interactive platform for information producers and consumers. The exchange service application such as Barter Services, Item Exchange, etc. is called homogeneous symmetric publish/subscribe system, which amis to allow as many users as possible engaged in matched and improves the profit of the system, and is increasingly causing concern of electric-commerce companies. Cloud computing has emerged as a new paradigm of computing platform recently, which covers lots of areas of computing. Cloud computing makes it possible for high efficiency and large scale computing platform. How to allow as many users as possible engaged in matching and make the publish/subscribe system extensible are key points of publish/subscribe system and the key technologies as well. There are two parts in the thesis. The first one is providing efficient algorithms for system optimization to allow as many users matched successfully as possible. The other one is proposing a parallel publish/subscribe system with excellent extension by the computing resources and technologies of cloud computing.For the problem of system optimization to improve the number of users engaged in matched, the thesis proposes greedy algorithm and heuristic algorithm by considering the satisfaction degree of users. There are only a few research achivement related to system optimization. The system optimal algorithm is NP-complete when the length of cycle match is equal to or greater than 3. How to find an approximate solution of system optimization in a limited time is an important. The thesis proposes greedy and heuristic algorithms considering system’s benefit and users’satisfaction degree. Greedy algorithm pays more attention to user’s satisfaction degree. Heuristic algorithm defines two reductions for the cycle with length 3, and gets the approximation result by reductions reducing the problem scale gradually. The result of the experiment shows that, whereas the execution time of the algorithms is similar, the number of users in the result set of heuristic algorithm is 9% to 24% more than that of greedy algorithm.For the problem of the extension of publish/subscribe system, the thesis designs a Twitter Storm-based parallel matching framework publish/subscribe. Despite the fact that there is a lot of research about the combination of cloud computing and publish/subscribe, there does not exist publish/subscribe platform based on Twitter Storm. The aim of the Twitter Storm-based publish/subscribe framework is to accelerate the matching between events and subscriptions for large scale information. First, to the lack of effective index in cloud environment, the thesis designs an index on HBase by dimension transform and space filling curve, and uses R-tree as the index structure making the query of event matching feasible. Second, the system adopts parallel matching processing event matching, which improves the matching speed effectively and hence improves the throughput of the system. At last, the experiment shows that the existence of index decreases the query time greatly and that the parallel framework has good scalability.