Nuclear Clustering Algorithm on State Grid's IT Operation Log

Abstract

With the continuous evolution of the IT infrastructure of State Grid Data Center and the growing operational data in the electric power system, how to quickly and automatically cluster the operation log in the data center of State Grid has become a key issue in the IT operation and maintenance of the data center. As the most commonly used algorithms in data mining, a clustering algorithm from data mining is adopted to handle the operational log data of State Grid IT data center, which can be used to effectively discover the changes of the topology structure during the operation of the IT infrastructure. Specifically, because the traditional sequential clustering algorithm lacks the ability to discover potential links in logs, this paper proposes a self-destructive nuclear clustering algorithm SDN-means, which aims at the business and data characteristics of the IT infrastructure system of State Grid data center, in order to effectively classify the operational log data of State Grid IT data center during the operation of State Grid. Through the analysis of the running logs of State Grid data center with obvious time series characteristics, the proposed SDN-means algorithm can effectively outperform the existing approaches on the operation of the data center of State Grid.