Design Of The Network Alarm Management System In Cloud Computing Environment

With the popularization of the computer network, the diversification of the terminal and the continuous progress of the data collection technology, the quantity of network information presents explosive growth trend. The traditional network management software based on relational database has many problems, such as large network bandwidth consumption, high management platform load, poor reliability and so on. Based on cloud computing environment, this paper focuses on the research of the On-Line Analytical Processing(OLAP) and association rules of network alarm data.Firstly, in view of the existing problems such as the single dimension, low efficiency, system memory overflow or even crash, the Hive-based alarm information OLAP method for the next generation network management application is proposed. This proposed method uses HBase to map the real-time collected massive network alarm data which is preprocessed into Hive. Then the N-D model and star model are constructed respectively to realize rollup and cube. Regardless of the presence or absence of inter domain devices the method is applicable. Experimental results show that the proposed method can not only realize the data bulk upload, but also reveal the hidden alarm location knowledge and some useful information of the users and systems from many dimensions, so as to provide decision support for network management.Secondly, an improved fast data mining algorithm based on MapReduce is proposed considering the characteristics such as massive, redundancy, relevance, etc. of the alarm data in the next generation network management application, as well as the double bottleneck problem of memory and computation time of the existing PFP algorithm. The structured storage of alarm data model is employed in the proposed algorithm. Furthermore, the repeated scanning of the entire database is reduced by simply scanning the data cube. In order to obtain the conditional mode, a shared path is adopted to reduce the number of tree traversal, and the MapReduce framework is applied to realize the parallel computing. Experimental results reveal that the proposed algorithm is efficient, scalable and reliable. The improved algorithm can find out the relationship between the states detected from the alarm system, reduce the duplication of alarm, and provide the basis for the equipment maintenance and network management.Finally, through integrating the proposed Hive-based alarm information OLAP method and the MapReduce-based fast data mining algorithm, an alarm management system for cloud computing environment is designed. Additionally, the modules of the network alarm management system are designed and implemented specifically, including data collection, storage, refresh, analysis, mining and data exchange, etc. The characteristics of the proposed system are that the analysis results of OLAP are the foundation of data mining. In addition, it can find more complex related information through data mining to expand the depth of OLAP, which makes the analysis intelligent, mining goal-oriented, so as to enhance the practical value of alarm management system.