A method for evaluating server health

Abstract

The invention provides a server health degree evaluation method which comprises the steps: S1: acquiring a network element indexes of a server, S2: acquiring an alarm level of each network element index and an index threshold of each alarm level according to index threshold monitoring information of a network manager, S3: acquiring a performance value of each network element index in real time, calculating a health degree reduction score of each network element index according to each alarm level, the index threshold of each alarm level and the performance value of each network element index when the performance value of each network element index does not exceed an upper index threshold limit or is not below a lower index threshold limit, and S4: calculating a health degree of the server, wherein the index threshold comprises the upper index threshold limit and / or the lower index threshold limit. The server health degree evaluation method can reflect the health degree of the server more scientifically, actually and effectively; and in addition, an evaluation result obtained by the method can help network management personnel to quickly find out the server more likely giving an alarm or giving a serious alarm.

Description

technical field [0001] The invention relates to the field of computers, in particular to a server health evaluation method. Background technique [0002] The work content of different servers is very different. Many servers often run continuously, coupled with unbalanced loads and other reasons, causing server failures to emerge in endlessly. The earlier server monitoring method is very simple: issue instructions to each server to collect relevant indicators to judge its operating status. However, with the rapid expansion of server scale, the method of judging faults by instructions for a single server has been unable to meet the demand. [0003] In order to effectively monitor a large number of servers, various server performance monitoring software appears on the market. These software can effectively monitor various indicators, such as cpu, memory, disk, network traffic, etc. By understanding the performance and status information of each monitoring indicator, you can ...

AI Technical Summary

Problems solved by technology

[0005] Most literatures use network element indicators (or components) to weight the overall health of the system, or use expert experience to determine the weight, or use the breadth-first traversal algorithm of the graph to determine the weight. However, the health results obtained in this way are very low. Difficult to combine directly with alarm analysis

No matter which weight determination method is used above, and regardless of whether the weight setting is accurate or not, the evaluation results are difficult to be directly used to assist alarm analysis due to the failure of the alarm analysis goal.

The evaluation results of server health need to directly determine the severity of the server in the alarm analysis, and the above weights are difficult to achieve this purpose

Images