Research For The Network Alarms Correlation Analysis And The Implementation And Optimization Of The System

With the popularity of the Network, the size of Computer Network is increasing sharply. And the structure of Computer Network becomes more complex, which makes the management and the maintenance for the Network becomes more difficult. The network fault management, which is one of the basic functions of the network management, is the basis for the five functions of network management.With the researching of some open source framework and some major software, which is widely used both in native and foreign (such as Netview made by IBM, OpenView made by HP and so on), and combining with the actual network environment and specific needs, the author chose the open source framework named Enterprise Java Bean 3.0 and the asynchronous messaging mechanism named Java Message Service to design and to implement a rule-based network fault management module.There are three modules included in this network fault management system: fault alarms collection, fault alarms correlation analysis and alarms processing. From the vision of architecture, it can be divided into client, server and database, these three components are splited from each other, only interacting through specific interfaces. Fault alarms collection mainly use the JMS technology which supports the synchronous message-driven and it can be separation with other monitoring modules by the message middleware in logic. In addition, the fault alarms collection modules also able to identify the incomplete or inconsistency data that transmitted through the network and the module can eliminate or correct the error data. Fault alarms correlation analysis module works for correlation the fault alarms to eliminate the unnecessary alarms. It based on the rules included: compression rule, overturn rule and the topology-based correlation rule. The alarms processing module works for submitting the correlation analysis results to the network administrator by E-mail and recording the results into the database.This system can eliminate the redundant and the error alarms, to help the network administrator to local the source fault quickly and easily. It can reduce the processing time for the network administrator to repair the network problem greatly. In performance, we reduce the database querying time greatly by the database regular backup plan and optimize the Java Persistence API . With the experiment, we show that this system has the high rate of depression among the fault alarms.