Distributed system monitoring method stepping over data center clusters and system

Abstract

The invention relates to a distributed system monitoring method stepping over a data center cluster and a system. The system has a multilayered distributed framework and comprises a business layer, a supporting layer and a monitoring framework; the monitoring framework consists of a business end, a control end, a collecting end and a monitoring end; the business layer comprises two function objects such as the business end and the control end; and the supporting layer comprises two function objects such as the collecting end and the monitoring end. The monitoring framework is managed in a unified manner, is opened stably, is used for cloud computing and steps over a plurality of data centers.

Description

technical field [0001] The present invention relates to the field of server cluster monitoring, and in particular to a monitoring system and method across multiple data centers. Background technique [0002] With the increasing application of cloud computing to various fields of the information industry, the number and scale of data centers on which cloud platforms are based has gradually increased, and businesses running on a single cloud platform span multiple data centers. How to effectively monitor and manage a large number of cluster devices across data centers and regions and ensure the high performance and high availability of the system is a necessary and urgent problem currently facing. [0003] The first monitoring architecture commonly used in the industry is pyramidal. For example, SuperMon, a general-purpose monitoring system software for large-scale Linux clusters developed by the National High Performance Computer Laboratory of a university, is divided into th...

AI Technical Summary

Problems solved by technology

[0005] (1) A large amount of monitoring data is collected from the lower-level server nodes to the general monitoring node at the same time. At the same time, the general monitoring node also responds to the request for monitoring data from the business side, which undoubtedly greatly increases the input and output pressure of the general monitoring node, and also directly leads to stability. Reduced performance, unable to meet daily business needs well

[0006] (2) The role of nodes in the monitoring system is unique. If there is a problem with the central monitoring node, part of the monitoring system will fail; if the main monitoring server node fails or goes down, the entire monitoring system will be paralyzed

[0007] (3) In the face of business across multiple data centers, it will be difficult for the total monitoring node to span multiple data centers under the bottleneck constraints of the total data volume and architecture, which will greatly slow down the response time and fail to meet real-time requirements

This invention creates a sub-center monitor for each sub-cluster, and pushes the collected internal information of the cluster to the central monitoring server, which still causes the problem of huge network and node I / O of the central monitoring server to be unsolved

It does not meet the personalized monitoring and customization needs of the business, and the dynamic scalability and runtime stability of nodes at each layer are not good

Images