Novel clustering algorithm based on timing alarm

Abstract

The invention discloses a novel clustering algorithm based on timing alarm, and mainly solves the clustering problem of massive timing alarm data generated by large-scale clustering equipment in a cloud environment in the prior art. According to the method, a new relation matrix between alarms is constructed by utilizing the relation of time differences between every two alarms in a set time window, and then K-is utilized; and the column vectors in the relation matrix are clustered by a K-means algorithm to obtain an alarm clustering result. By means of the scheme, the purposes that operationand maintenance personnel usually utilize clustering results of the alarm data to accurately locate, troubleshoot and repair faults and errors and maintain normal operation of the system are achieved,and the novel clustering algorithm has high practical value and popularization value in the field of operation and maintenance of cloud service providers.

Description

technical field [0001] The invention relates to the technical field of cloud computing, in particular to a novel clustering algorithm based on time series alarms. Background technique [0002] With the rapid development of cloud computing, there are more and more devices in the cloud environment, and the topological structure between devices is becoming more and more complex. The difficulty of device fault diagnosis and troubleshooting also increases. For example, the occurrence of an alarm often leads to a series of alarms, and continuous iteration leads to a large number of alarms in the system, and the alarms generated by different reasons are intertwined. Affected by a large number of other irrelevant alarm information, it is impossible to accurately locate, troubleshoot and repair the fault. [0003] In the recovery process of large-scale websites, fault location and troubleshooting occupy about 93% of the total time. It can be seen that maintenance personnel must ana...

AI Technical Summary

Problems solved by technology

The above methods all use mining based on association rules and frequent patterns. These mining algorithms rely on high support and high confidence conditions, and do not take into account the existence of new relationships in the time of various alarms within the time window.

However, NingPing and Cui Yun are responsible for analyzing the technology and tools of intrusion alarms. The method needs to define a semantic relationship library in advance, which is not time-sensitive.

Images