System and method for detecting health degree of application cluster

Abstract

The invention relates to a system and a method for detecting the health degree of an application cluster. The method comprises the following steps that to-be-monitored application cluster information is initialized; the data acquisition unit acquires performance data generated by the APM tool and stores the performance data to a local database; the baseline calculation unit obtains a dynamic baseline according to historical time sequence performance data in a local database; the anomaly detection unit is used for carrying out anomaly detection on the real-time performance data in the local database; and the alarm unit judges the health degree of the application cluster according to the anomaly detection result of the anomaly detection unit and sends out alarm information. Compared with the prior art, based on the data acquisition unit, the dynamic baseline calculation unit, the anomaly detection unit, the alarm unit, the task scheduling unit and the local database, the APM data of the application cluster is effectively stored, the health degree of the application cluster is obtained, and the accuracy and efficiency of health degree detection of the application cluster are improved; and the reliability of application cluster health degree detection is improved.

Description

Technical field[0001]The present invention relates to the application of cluster health detection, in particular, to a system and method for detecting application clustering health.Background technique[0002]With the popularity of micro service architecture, more and more IT projects are developed using micro-service architecture. It is true that micro-service achieves the decoupling between modules, and meets the requirements of agile development of enterprises with its independent deployment, fast iteration, etc. However, as the service module is increasing, the call between the service is increasing, and we urgently need to apply service call analysis (hereinafter referred to as APM) tool to monitor the call of each application service and response performance.[0003]At present, more common APM tools have Naver's PinPoint, Google's Dapper, Twitter's zipkin, Taobao's eagle, and public comment CAT, etc.[0004]Although the above tools can show the performance of each service and the ca...

AI Technical Summary

Problems solved by technology

[0005] 1. The setting of alarm thresholds for some indicators depends on user experience and cannot be set scientifically;

[0006] 2. According to changes in business pressure at different points in time, some performance indicators will show a dynamic trend, and existing tools lack the ability to establish a dynamic baseline for this indicator;

[0007] 3. Conventional threshold alarms have a low tolerance to faults and a high false alarm rate

Images