Method, apparatus, and storage medium for diagnosing failure based on a service monitoring indicator of a server by clustering servers with similar degrees of abnormal fluctuation

Patent metadata

Field	Value
Publication number	US-10805151-B2
Application number	US-201916354074-A
Country	US
Kind code	B2
Filing date	Mar 14, 2019
Priority date	May 31, 2018
Publication date	Oct 13, 2020
Grant date	Oct 13, 2020

Abstract

A method, an apparatus and a storage medium for diagnosing failure based on a service monitoring indicator are provided. Service monitoring indicator of a server is analyzed to obtain a degree of abnormal fluctuation in the service monitoring indicator. Servers with similar degrees of abnormal fluctuation are clustered according to the degree of abnormal fluctuation of the service monitoring indicator, to obtain clustered results that include the servers and the service monitoring indicator. A location where the service fails is determined according to the clustered results.

First claim

What is claimed is:

1. A method for diagnosing failure based on a service monitoring indicator, wherein the method comprises:
analyzing, according to an abnormality fluctuation detection algorithm, a service monitoring indicator of a server, in which a service is located, to obtain a degree of abnormal fluctuation in the service monitoring indicator;
clustering servers with similar degrees of abnormal fluctuation according to the degree of abnormal fluctuation of the service monitoring indicator, wherein clustered results comprise the servers and the service monitoring indicator; and
determining, according to the clustered results, a location where the service fails;
wherein the analyzing, according to an abnormality fluctuation detection algorithm, a service monitoring indicator of a server, in which a service is located, to obtain a degree of abnormal fluctuation in the service monitoring indicator, comprises:
calculating a degree of first fluctuation and a degree of second fluctuation of the service monitoring indicator using kernel density estimation (KDE), wherein the degree of first fluctuation is a degree of fluctuation in the service monitoring indicator on the same day that the failure occurs, and the degree of second fluctuation is a degree of fluctuation in the service monitoring indicator within a preset time period before the failure occurs; and
comparing, according to an extreme value theory, the degree of first fluctuation and the degree of second fluctuation to determine a probability that the degree of first fluctuation is an abnormal fluctuation, so as to obtain the degree of abnormal fluctuation of the service monitoring indicator.

2. The method according to claim 1 , wherein the clustering servers with similar degrees of abnormal fluctuation according to the degree of abnormal fluctuation of the service monitoring indicator comprises:
clustering, according to the degree of abnormal fluctuation in the service monitoring indicator, servers with similar degrees of abnormal fluctuation using a preset clustering algorithm, wherein the preset clustering algorithm is a hierarchical clustering algorithm or a Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm.

3. The method according to claim 2 , wherein, before the clustering servers with similar degrees of abnormal fluctuation using the preset clustering algorithm, the method further comprises:
for each server, sorting the degree of abnormal fluctuation in the service monitoring indicator corresponding to the each server to obtain a sorted result corresponding to the each server;
calculating a correlation coefficient between sorted results corresponding to two servers; and
correspondingly, the clustering servers with similar degrees of abnormal fluctuation using the preset clustering algorithm comprises: clustering servers with similar degrees of abnormal fluctuation using the preset clustering algorithm based on the correlation coefficient.

4. The method according to claim 2 , wherein each computer room comprises a plurality of the servers, the service comprises at least one module, and the clustering servers with similar degrees of abnormal fluctuation using the preset clustering algorithm comprises:
clustering the servers comprised in the each computer room according to the module comprised in the service.

5. The method according to claim 1 , wherein the determining, according to the clustered results, the location where the service fails comprises:
sorting the clustered results according to a preset sorting strategy; and
determining, according to the sorted results, a location where the service fails.

6. The method according to claim 1 , wherein, before the analyzing, according to an abnormality fluctuation detection algorithm, a service monitoring indicator of a server, in which a service is located, to obtain a degree of abnormal fluctuation in the service monitoring indicator, the method further comprises:
obtaining historical data of the service monitoring indicator of a server in which the service is located; and
correspondingly, the analyzing, according to an abnormality fluctuation detection algorithm, a service monitoring indicator of a server, in which a service is located, to obtain a degree of abnormal fluctuation in the service monitoring indicator comprises: analyzing, according to an abnormality fluctuation detection algorithm, the historical data to obtain the degree of abnormal fluctuation in the service monitoring indicator.

7. An apparatus for diagnosing failure based on a service monitoring indicator, the apparatus comprising a processor and a computer-readable medium for storing program codes, which, when executed by the processor, cause the processor to:
analyze, according to an abnormality fluctuation detection algorithm, a service monitoring indicator of a server, in which a service is located, to obtain a degree of abnormal fluctuation in the service monitoring indicator;
cluster servers with similar degrees of abnormal fluctuation according to the degree of abnormal fluctuation of the service monitoring indicator, wherein clustered results comprise the servers and the service monitoring indicator; and
determine, according to the clustered results, a location where the service fails;
wherein the program codes further cause the processor to calculate a degree of first fluctuation and a degree of second fluctuation of the service monitoring indicator using kernel density estimation (KDE), wherein the degree of first fluctuation is a degree of fluctuation in the service monitoring indicator on the same day that the failure occurs, and the degree of second fluctuation is a degree of fluctuation in the service monitoring indicator within a preset time period before the failure occurs; and
compare, according to an extreme value theory, the degree of first fluctuation and the degree of second fluctuation to determine a probability that the degree of first fluctuation is an abnormal fluctuation, so as to obtain the degree of abnormal fluctuation of the service monitoring indicator.

8. The apparatus according to claim 7 , wherein the program codes further cause the processor to:
cluster, according to the degree of abnormal fluctuation in the service monitoring indicator, servers with similar degrees of abnormal fluctuation using a preset clustering algorithm, wherein the preset clustering algorithm is a hierarchical clustering algorithm or a Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm.

9. The apparatus according to claim 8 , wherein the program codes further cause the processor to:
before the clustering servers with similar degrees of abnormal fluctuation using the preset clustering algorithm,
for each server, sort the degree of abnormal fluctuation in the service monitoring indicator corresponding to the each server to obtain a sorted result corresponding to the each server;
calculate a correlation coefficient between sorted results corresponding to two servers; and
correspondingly, when clustering servers with similar degrees of abnormal fluctuation using the preset clustering algorithm, the program codes further cause the processor to: cluster servers with similar degrees of abnormal fluctuation using the preset clustering algorithm based on the correlation coefficient.

10. The apparatus according to claim 8 , wherein each computer room comprises a plurality of the servers, the service comprises at least one module;
when clustering servers with similar degrees of abnormal fluctuation using the preset clustering algorithm, the program codes further cause the processor to:
cluster the servers comprised in the each computer room according to the

Assignees

• Beijing Baidu Netcom Sci & Tech Co Ltd

Inventors

• Liu Ping
• Chen Yu
• Chen Yun
• Qu Xianping

Classifications

• H04L41/0695 Primary

  the faulty arrangement being the maintenance, administration or management system · CPC title

• H04L41/142 Primary

  using statistical or mathematical methods · CPC title

• H04L41/0853

  by actively collecting configuration information or by backing up configuration information · CPC title

• H04L41/0677 Primary

  Localisation of faults · CPC title

• H04L41/064

  involving time analysis · CPC title