Method and device for assessing capacity of distributed cluster processing specified service

Abstract

The invention provides a method and device for assessing the capacity of a distributed cluster processing a specified service. The method comprises the following steps that performance indexes for assessing the capacity of the distributed cluster processing the specified service and resource parameters, constraining the performance indexes, of the distributed cluster are determined; performance index data and resource parameter data are acquired; a relation model of the performance indexes and the resource parameters is determined by adopting a machine learning algorithm by means of the performance index data and the resource parameter data; and the capacity of the distributed cluster processing the specified service is determined according to the relation model. According to the technicalscheme, the influences possibly caused on the services operating on line can be avoided, the corresponding service processing capacity of the distributed cluster operating on line can be assessed onthe basis of the practical service data, the condition that the service serviceability of the distributed cluster operating on line is influenced can be avoided, and the advantage of reducing the assessment cost is achieved.

Description

technical field [0001] The present application relates to a method for evaluating the processing capability of a distributed cluster, in particular to a method and a device for evaluating the capability of a distributed cluster for processing specified services. Background technique [0002] For any system that uses a large-scale distributed cluster to provide business services, the processing capacity of the business services it provides is limited. When the maximum upper limit of its processing capacity is reached, the system cannot continue to provide corresponding business services. [0003] In practical applications, it is necessary to ensure the high availability of the services provided by the services deployed on the distributed cluster, that is, to ensure that the processing capacity that can provide corresponding services within a controllable time is greater than the demand for the services. [0004] In different business development stages, the number of busines...

AI Technical Summary

Problems solved by technology

[0012] Among the above methods, the "simulated request stress test" can only initiate a single logical request in the test environment, and cannot fully cover complex online requests; the "online traffic drainage stress test" needs to build a completely consistent online The standard test environment, including the upstream and dependent downstream services that the original service depends on, requires expensive pressure testing costs, especially in a distributed cluster environment with multiple service entities; "Modify the weight pressure test method "First of all, relying on cluster support to dynamically adjust load balancing is a good method for a simple CS architecture, but in a distributed cluster with a scattered organizational structure, there are fewer business scenarios that can be analyzed and covered, and the implementation cost is relatively high.

[0013] To sum up, the existing methods of stress testing the business processing capabilities of distributed clusters have the problems of high requirements for the test environment and high test costs

Images