Evaluation And Research On Node Failure Processing Capability Of Cloud Platform

In recent years,cloud computing technology has been matured,companies have either built their own cloud platforms or migrated applications to the cloud.With its on-demand service and flexible scalability,cloud computing solves the problems of business peaks and big data processing that are difficult to adapt to traditional computing models,providing users with almost unlimited infrastructure resources.However,while cloud computing brings convenience to users,some cloud failure events also cause users to have great concerns about the security of the cloud.As a large-scale distributed system,the cloud platform consists of a large number of physical nodes.These physical nodes may fail at any time.In order to reduce the impact of cloud resource failures on user services,cloud service providers need to establish cloud platforms with failure processing capabilities.However,it is not enough to establish a fault-tolerant cloud platform.It is also necessary to perform fault tolerance verification on the cloud platform.There are a lot of researches on how to build a highly available cloud platform,but there are few studies related to cloud platform fault tolerance assessment.From this perspective,this paper studies the node failure processing capability of the cloud platform.The proposed cloud platform node failure processing capability evaluation method can be used as a disaster drill of the measured cloud to test the disaster recovery performance of the cloud platform.The evaluation result can help the cloud service provider to verify the fault tolerance performance of the cloud platform or find fault tolerance bugs of the cloud,so the cloud administrators can solve those defects in advance to avoid losses due to the vulnerability in the actual production environment.This paper first analyzes the high-availability technology and high-availability architecture of the cloud platform,investigates the common deployment structure of the cloud platform,and proposes a common node failure evaluation system which is suitable for different cloud platforms.The system includes the cloud platform fault injection method,Evaluation indicators,testing procedures and testing load.As for the proposal of fault injection method,this paper analyzes and summarizes the causes of the failure of the cloud failure event in recent years,and selects four types of faults: CPU,memory over-occupancy fault,network fault,power failure fault.For network faults,this paper analyzed different types of network traffic in the cloud platform,and designed different types of network fault injection for different nodes.Then,from the perspective of cloud service providers and users,this paper proposes evaluation indicators for evaluating cloud platform node failure processing capabilities,and proposed different fault injection and testing load for different nodes.Based on the proposed evaluation method,a detailed test flow is designed.Finally,based on the proposed evaluation system,this paper designs and implements the Automatic Test Evaluation Tool.The tester can run the test tool by filling in the test item configuration information.After the test,the test tool will generate a visual chart,which is convenient for the tester to intuitively see the performance changes of the cloud platform before and after fault injection.Finally,this paper selects the open source Open Stack cloud operating system to build an experimental cloud platform with fault-tolerant processing capability,and uses the evaluation tool of this paper to test and analyze and evaluate the test results.The experiment proves that the node failure evaluation system proposed in this paper can effectively test and verify the fault tolerance of the cloud platform.And the Failure Test evaluation tool developed in this paper can effectively measure the node failure processing capability of the cloud under test.