| With the rapid development of information technology and business intelligence, the role information systems played in corporate decision-making is more and more important. Serving as the important support of enterprise, traditional information system has many shortcomings. Such as the scalability of the system is poor, the software development process is contained in business logic, and the ability to respond to changes of the enterprise market is weak. However, the information system based on business rules has the features of excellent adaptability and flexibility, exactly making up the shortcomings of traditional information system. In addition to the scalability and flexibility, efficiency is another factor of intelligence system which needs to be considered. Compared with the traditional centralized computing, which the computing almost on just one computer, distributed computing finishes the computing of large amount of calculation on computer cluster. Distributed computing greatly improves the efficiency of information process and the utilization of computer resources, making up the shortcomings of centralized computing on low computational efficiency and limited host processing capability. Therefore, based on current demand for business information systems, the article puts forward solutions for the realizing of distributed rule matching system which based on MapReduce method.This paper firstly researches the traditional rules engine which represented in Drools, and analyzes the applications, advantages and disadvantages of Drools rules engine. Then the author researches the "Map" and "Reduce" data processing of MapReduce method, and making it as the theoretical basis of the design of distributed rules processing framework and the optimization.of facts processing. Then the author puts forward one distributed rules processing framework which based on MapReduce method. The proposed framework is based on the characteristics of parallel computing and the idea of MapReduce method, which is "Divide and Conquer". As to the circumstance that may arise under the MapReduce method, which dataflow excessively concentrate on Reduce node, the author introduces the balancing strategy between "Map" and "Reduce". By Dixon assay, we can judge the data flow on multiple Map nodes is whether equivalent or not. Then the balancing strategy is triggered only when necessary. When the balance strategy is triggered, the data flow and sub-rules are shifted between Map nodes to achieve the facts’distribution balance on Map nodes. And then after the strategy is triggered, the calculation on every Reduce node will be balanced, and the computing resource is also fully used. Finally, the author designs and realizes the credit card aid-application system based on MapReduce method through the software engineering method, which including requirements analysis, design, implementation, and testing process. The design and realize of the credit card aid-application system is based on the distributing rules processing framework put forward before. Then the author compares the credit card aid-application system with the system prototype which realized based on Drools to verify the efficiency of distributed rules processing. |
