Tail delay evaluation system and method for a cloud platform

Abstract

The invention belongs to the technical field of computer applications, and provides a tail delay evaluation system and method of a cloud platform. The tail delay evaluation system includes a server, client and logic analysis module; the client includes a request generation module and a delay statistics module; the request generation module generates random requests at a fixed rate, and the client sends the generated requests to the deployment through the TCP / IP protocol On the server side on the cloud platform, the server feeds back the processing results to the client to complete a request. The logic analysis module reads the delay statistics file, and obtains the linear regression equation between the processing quantity in the time period and the time period sequence. The present invention realizes that the tail delay probability distribution diagram that can more accurately describe the tail delay performance between cloud platforms can be obtained only by monitoring at the client end. This method can more accurately describe the performance of the tail delay performance through a small number of tests combined with the calculation of probability statistics.

Description

technical field [0001] The invention belongs to the technical field of computer applications, and in particular relates to a system and method for evaluating the tail delay of a cloud platform. Background technique [0002] With the development of cloud computing, more and more companies choose to deploy large-scale, high-complexity, and high-density computing processes on rented cloud platforms in a short period of time. However, due to the von Neumann architecture Inherent defect, when multi-user virtual machines are allocated on the same physical machine, the overall performance will have great instability and obvious performance degradation, which are mainly due to the following two reasons: [0003] On the one hand, when requests from different users work on the same physical machine, the underlying hardware cannot identify the specific source user of the instruction request, which leads to a large number of virtual machine users running on the same physical machine. I...

AI Technical Summary

Problems solved by technology

[0005] Tail latency is defined as the 90th, 95th, or 99th percentile latency of all latency-sensitive requests under such performance degradation and random disturbances, which tend to be much higher than the average latency, while For delay-sensitive services, these tail delays lead to very poor user experience, so how to correctly evaluate the performance of cloud platforms in terms of tail delay performance has always been an important issue of concern to the enterprise community

[0006] On the one hand, the existing test tools need to modify both the client and the server to a certain extent, leaving the corresponding test interface, and also need the root authority of the server, and many times the user does not have the corresponding root authority. On the other hand, the existing The test tool only performs simple measurement when evaluating the tail delay, the data obtained by the short-term measurement will have large fluctuations, and the long-term measurement is difficult to ensure the stability of the test environment between different platforms

Images