The Research Of Key Technologies And Applications Of Alarm Correaltion In Optical Transport Networks

The optical transport networks have grown considerably in size, complexity and bandwidth. It is very hard to manage so complicated networks, especially in fault management aspect. In order to locate faults, it is possible to correlate alarm messages that are deemed to have been produced because of a common underlying fault. The alarm correlation is a conceptual interpretation of multiple alarms such that new meanings are assigned to these alarms. So the multiple alarms are replaced by one meaningful alarm message. It is a generic process that underlies different network management tasks such as alarm filtering, network fault diagnosis, generation of corrective actions, proactive maintenance and network behavior trend analysis.Based on the related internal and external researches, this thesis systematically focuses on the academic researches and the practical development of alarm correlation in optical transport networks. The alarm pretreatment method and the fault diagnosis method are proposed for the alarm correlation analysis in the same sub-network. In addition, based on Bayesian networks the fault diagnosis method is also proposed for the fault diagnosis among different layer networks. The main achievements are as follows in thesis.Firstly, the thesis proposes the the alarm pretreatment process, which is one of important parts in alarm correlation. The alarm pretreatment process is divided into two parts, including alarm filtering and alarm compression. The alarm filtering process is responsible to fetch the key alarm groups form the large numbers of alarm series caused network faults, and filter other meaningless and redundant alarms. The alarm compression process is responsible to compress the same key alarms in order to reduce the total number of key alarms in the groups. Meanwhile a method is proposed to obtain the alarm observation window.Secondly, the process of fault diagnosis is the other of important parts in alarm correlation. The thesis proposes the fault diagnosis process in the sub-networks, which combines dependency graph among communication entities with causality graph between alarms and faults. The fault diagnosis process is divided into there parts, including fault laocation, alarm sequence search and fault diagnosis. The fault location sub-process is responsible to locate the fault sites with dependency graph and set up the key alarm sub-groups. One sub-group contains all the key alarms reported in one fault site. The alarm sequence search sub-process is responsible to search the typical alarm sequence defined in the rules from the key alarm sub-groups and set up the related alarm groups. The fault diagnosis sub-process is responsible to match the related alarm groups to the fault rules. According these sub-processes, the final faults in the sub-network are diagnosed at last.Thridly, the thesis proposes the fault diagnosis process among different layer networks based on Bayesian network, in order to settle the uncertainty in the root fault diagnosis. At first, the thesis proposes the hierarchy structure of communication networks and constructs the dependency graph among layer networks. The dependency graph among layer networks only focuses on the dependency relation between two layer networks which hiddens the internal complex structure in the layer networks. Then the Bayesian network is constructed from the dependency graph. And the probability in the Bayesian nodes are obtained from the expert experience and history statistical information. After the Bayesian network is set up, the fault diagnosis problem is transformed to Bayesian reasoning.Finally, the thesis discusses the framework and key technologies for alarm correlation realization in optical transport networks, including the knowledge requirement and representation, system database design, alarm correlation processes and user interfaces.