A Study On Clustering Fusion And Correlation Analysis Of Alarm Information In Heterogeneous Networks

With the development of heterogeneous networks, the scale and the structure of telecommunication networks have become more and more complicated. The difficulty in dealing with network alarms caused by this situation is also increasing as well. Therefore, an efficient and flexible alarm management scheme is becoming increasingly significant in heterogeneous networks. In this paper, we have attached our emphasis on two aspects of alarm management and proposed the corresponding algorithm respectively to each specific one. The first aspect is the work on alarm gathering and the second aspect is related to alarm analysis. Combining with both of the two aspects in alarm management, we finally come up with corresponding algorithms to solve these problems in alarm gathering and alarm analysis, respectively, so as to achieve the target for an effective alarm management.For the issue of alarm gathering in heterogeneous networks, we propose a corresponding scheme for alarm clustering and fusion. First of all, through relevant preprocessing, we achieve the uniform alarm frame format and the alarm similarity matrix in heterogeneous networks. Then, we employ the fuzzy clustering method so as to make a clustering analysis for similar alarms. Based on the distance of alarms to the root fault, we can gather the similar alarms into a same fuzzy alarm cluster. Furthermore, according to the Dempster-Shafer theory with consideration of confidence, we further make a fusion among these alarms in a same cluster and obtain the corresponding comprehensive alarms, which can significantly reduce the data volume in alarm processing.For the issue of alarm analysis in heterogeneous networks, we also propose an effective scheme to handle it, including the distributed analytical model and the distributed correlation algorithm. Considering the structure of heterogeneous networks, the distributed model we proposed in this paper does have a lot of potential advantages, especially for the small number of the candidate sets in alarm database. This kind of superiority can greatly help us enhance the efficiency of alarm correlation, so as to largely reduce the time consumption in correlation analysis under the environment of heterogeneous networks. The improvement achieved by our distributed scheme will certainly bring about a solid support for the better alarm management in heterogeneous networks.