CN112559129B - Device and method for testing load balancing function and performance of virtualization platform - Google Patents

Abstract

The invention discloses a device and a method for testing the load balancing function and performance of a virtualization platform. The method comprises the following steps: the environment configuration module reads the test configuration file and initializes the test environment; the data acquisition module acquires and stores basic performance data of the test cluster; the load sending module sends a task request to the load balancer; and the analysis and evaluation module calculates the load balancing performance and generates a test report. The method extracts the performance test indexes from multiple angles by using the basic performance data of the virtualization platform test cluster when receiving the load, determines the weight of each performance test index by using an entropy method, finally obtains the virtualization platform load balancing performance score, and realizes the comprehensive evaluation of the virtualization platform load balancing function.

Description

Test device and method for load balancing function and performance of virtualization platform

technical field

The invention belongs to the field of physical technology, and further relates to a testing device and method for the load balancing function and performance of a virtualization platform in the field of automatic testing technology. The invention can be used for testing the load balancing function and performance of the virtualization platform.

Background technique

In recent years, cloud platforms have gradually become the primary choice for software vendors to deploy services. The underlying implementation of the cloud platform is based on virtualization technology, so the cloud platform is a specific instance of the virtualization platform, and the virtualization platform is the generalization of the cloud platform. The virtualization platform is a platform for resource sharing, on-demand consumption, dynamic expansion, and load balancing. For this platform, the load balancing function is one of the main functions it provides, which involves two roles, namely the virtualization platform service provider and the virtualization platform service buyer. For virtualization platform service providers, the load balancing function is an important research direction. In order to achieve better performance of the load balancing function, the virtualization platform service provider will continuously optimize the algorithm of its load balancing function; for virtualization platform service buyers, It needs to compare the load balancing functions of different virtualization platforms according to business requirements, so as to select the most suitable virtualization platform load balancing service.

To sum up, a reasonable load balancing testing method is very important. However, the current testing methods for virtualization service platforms at home and abroad focus more on the evaluation of its overall performance; the evaluation of the load balancing function of the virtualization platform is only in a relatively simple testing stage. Therefore, an objective and fair load balancing test method is needed to evaluate the quality of the load balancing function.

Suzhou Inspur Intelligent Technology Co., Ltd. disclosed a load balancing feature of virtualization platform in its patent document "Testing method, device, system, and equipment of load balancing feature" (patent application number 201910290965.1 application publication number: CN109981419 A). Automated testing methods and devices. The device disclosed in this patent document includes a request receiving module, a script uploading module, and a test result analyzing module. The request receiving module is used to receive the test request to the load balancing characteristic test, and builds a test cluster; the script upload module is used to upload the test script to the test cluster, and run the script in the cluster to obtain the test log of each node of the cluster; The test result analysis module includes a load value determination unit and a result determination unit. After receiving the test log, it determines whether the load balancing characteristic test is successful or not according to the determined load value. The disadvantage of this device is that: since the test result analysis module directly obtains the test results of each node through the script upload module, the obtained test data cannot be stored persistently, which is not conducive to testers when testing the load balancing function of other platforms. Compare and evaluate different load balancing functions based on the data. The steps of the automated testing method for testing the load balancing feature are: first, the management node determines the configuration file according to the user's test request, including the test preset threshold, the size of the test cluster, the load type, and the size of the load, and builds a test cluster according to the configuration file; Then upload the test script of the test load balancing feature to the test cluster, make the test cluster run the test script, and according to the node load situation, evenly distribute the load request to obtain the load log of each node; finally receive the test log of the test cluster, Using the final load status of each node recorded in the test log, it is judged whether each node is load balanced. The disadvantage of this automated testing method is that the tester needs to manually configure the predetermined threshold of load balancing, and the predetermined threshold is a key parameter to determine whether the test of load balancing characteristics is successful, so the test will be greatly disturbed by human factors; Whether the difference between the load value and the minimum load value is greater than the predetermined threshold to determine whether the load balancing feature meets the requirements can only enable testers to determine whether the load balancing service of the virtualization platform is available, but the load balancing service is available on various virtualization platforms Under such circumstances, it is difficult to further compare the quality of load balancing services of different virtualization service platforms, and it is impossible to choose the most suitable load balancing service.

In their paper "Performance Evaluation and Analysis of Load Balancing Alogorithms in Cloud Computing Environments" (International Conference on Web Research, pp145-pp151, 2016), MR Mesbahi et al. mentioned a method for testing the performance of cloud platform load balancing algorithms. . The steps of the test method are: firstly configure the traffic routing strategy between the user database and the data center in the cloud platform simulator, including the nearest data center, optimize the response time, and dynamically reconfigure the load; Choose different load balancing strategies on VMLoadBalancer, a component that simulates data center load scheduling, including three strategies: polling, throttling, and evenly distributing the current execution load; after configuring the above two steps according to actual test requirements, you can get up to nine common The cloud platform load balancing strategy combination, each test sends the same type and size of load to the cloud platform simulator according to the user test requirements, and obtains the response time of each node; finally calculates the average response time and cost of each algorithm respectively , so as to compare and evaluate each load balancing algorithm. The shortcomings of this test method are: the selected performance test indicators are too simple, and only the average response time and cost overhead are used to evaluate the load balancing function. When the above factors are not the main factors that testers need to consider, it is difficult for testers to The test results compare the quality of the load balancing service; each performance test indicator is compared separately, and the test indicators are not aggregated to form an overall load balancing performance evaluation model and evaluation system, and it is impossible to comprehensively evaluate the load balancing function of the virtualized service platform.

Contents of the invention

The purpose of the present invention is to solve the problems existing in the above-mentioned prior art, and propose a test device and method for the load balancing function and performance of a virtualization platform, which is used to solve the problem that the existing methods mainly focus on the basic performance of the load balancing function of the virtualization platform It is difficult to provide a more objective test of the virtualization platform load balancing function in the actual production and use process of virtualization platform service providers and virtualization platform service buyers. technical support questions.

The idea of realizing the purpose of the present invention is to start from the perspective of virtualization platform service providers and virtualization platform service buyers paying attention to the virtualization platform load balancing function in the actual production environment, combining the characteristics of the virtualization platform and the existing load balancing performance Based on the advantages of the test technology, four performance test indicators that can indirectly reflect the load balancing capability of the virtualization platform are proposed. The performance test indicators can be calculated from the basic performance data of the test cluster when receiving the load, and the performance test indicators are determined by the entropy method The weight of each indicator in .

To achieve the above object, the device of the present invention includes an environment configuration module, a data collection module, a data storage module, a load sending module and an analysis and evaluation module.

The environment configuration module is used to read the test configuration file; by interacting with the authentication service of the virtualization platform, obtain the management authority of the virtualization platform, and then interact with the virtualization platform through the authority to create a load balancer; according to In the configuration file, the test cluster virtual machine scale, the number of test cluster virtual machines, the test cluster virtual machine image, call the interface provided by the virtualization platform, create multiple virtual machines as a test cluster, and add the test cluster IP address to the load balancer in the load balancing pool;

The data acquisition module is used to test the benchmark performance data of the test cluster. When the task request is not sent, the basic performance data of the test cluster is collected by the virtualization platform monitoring system as the benchmark performance data of the load balancing function; in the load balancer When receiving a task request, collect the basic performance data of the test cluster through the monitoring system of the virtualization platform;

The data storage module is used to store the measured benchmark performance data in a database for persistent storage;

The load sending module is used to generate a request of HTTP GET type according to the total number of task requests in the configuration file, the type of task requests, and the percentage of different types of task requests in the total number of task requests, through the preset task request type and url address One-to-one mapping relationship, modify the url address in the HTTP GET request, continue to build different types of task requests, and send the built task requests to the load balancer; count the number of each type of task requests that have been built, If the percentage of the number of task requests of this type to the total number of task requests in the configuration file satisfies the percentage of the number of task requests of this type in the total number of task requests in the configuration file, stop building this type of request; determine whether the number of load sending rounds is equal to the preset in the configuration file If so, go to the next step to calculate the load balancing performance, otherwise, add 1 to the value of the number of load sending rounds and repeat the load sending step;

The analysis and evaluation module is used to calculate the performance loss of each test round test cluster when receiving a task request; calculate the responsiveness of each test round test cluster when receiving a task request; calculate each test round test cluster The degree of resource balance when receiving task requests; calculate the invalid request rate of the total time block of the test cluster for each test round when receiving task requests; according to the number of weight test rounds in the configuration file, select the number of weight test rounds For the corresponding performance test index, the weight of each index in the performance test index is determined by the entropy method; the maximum number of test rounds minus the number of weight test rounds is used as the remaining number of test rounds, and the performance corresponding to the remaining number of test rounds The test indicators are weighted and summed by the arithmetic mean value to obtain the load balancing performance score of the virtualization platform; the test report of the load balancing function test of the virtualization platform is generated on the front-end virtualization platform monitoring page.

Method of the present invention comprises the steps:

(1) The environment configuration module reads the test configuration file;

(2) Initialize the test environment:

(2a) The environment configuration module obtains the management authority of the virtualization platform by interacting with the authentication service of the virtualization platform, and then interacts with the virtualization platform through the authority to create a load balancer;

(2b) The environment configuration module calls the interface provided by the virtualization platform according to the test cluster virtual machine scale, the number of test cluster virtual machines, and the test cluster virtual machine image in the configuration file, creates multiple virtual machines as test clusters, and transfers the test cluster The IP address is added to the load balancing pool of the load balancer;

(3) Collect benchmark performance data of the storage test cluster:

The data acquisition module tests the benchmark performance data of the test cluster. When no task request is sent, the basic performance data of the test cluster is collected through the virtualization platform monitoring system as the benchmark performance data of the load balancing function. The data storage module stores the measured benchmark performance The data is stored in the database for persistent storage;

(4) Send a task request to the load balancer:

The load sending module generates HTTP GET type requests according to the total number of task requests, task request types, and the percentage of different types of task requests in the total number of task requests in the configuration file, and uses the one-to-one mapping relationship between preset task request types and url addresses , modify the url address in the HTTP GET request, continuously construct different types of task requests, and send the constructed task requests to the load balancer; separately count the number of constructed task requests of each type, if the task requests of this type If the percentage of the total number of task requests in the configuration file meets the percentage of the total number of task requests of this type in the configuration file, then stop building this type of request;

(5) Collect basic performance data when the storage test cluster receives task requests:

The data acquisition module collects the basic performance data of the test cluster through the virtualization platform monitoring system when the load balancer receives the task request; the data storage module stores the measured basic performance data into the database for persistent storage;

(6) The load sending module judges whether the number of load sending rounds is equal to the preset maximum number of test rounds in the configuration file, if so, execute step (7), otherwise, add 1 to the value of the load sending rounds and then execute step (4) ;

(7) Calculate load balancing performance:

(7a) According to the following formula, the analysis and evaluation module calculates the performance loss of each test round when the test cluster receives task requests:

Among them, l _p represents the performance loss of the test cluster in the pth test round when receiving the task request, n _p represents the duration of the task request received by the test cluster in the pth test round, and i represents the test cluster in the pth test round Receive the serial number in the time series of the task request duration, b represents the benchmark performance of the test cluster, and the value of the benchmark performance is the CPU utilization rate, memory utilization rate, network inflow bandwidth utilization rate, and network outflow bandwidth in the benchmark performance data The average value of the utilization rate, q _pi represents the actual performance of the test cluster at the i-th second of the p-th test round, and the value of the actual performance is the CPU utilization at the i-th second of the p-th test round in the basic test data, The average value of memory utilization, network inflow bandwidth utilization, and network outflow bandwidth utilization, the smaller the value of _lp , the better the load balancing performance of the virtualization platform;

(7b) According to the following formula, the analysis and evaluation module calculates the responsiveness of each test round test cluster when receiving task requests:

Among them, r _p represents the responsiveness of the p-th test round test cluster when receiving task requests, e represents a natural constant, and D _p represents the number of task requests that the p-th test round test cluster loses packets when receiving task requests, Q represents the total number of task requests constructed by the load sending module, the smaller the value of r _p , the better the load balancing performance of the virtualization platform;

(7c) According to the following formula, the analysis and evaluation module calculates the resource balance degree of each test round test cluster when receiving task requests:

Among them, d _p represents the resource distribution balance degree of the p-th test round test cluster when receiving task requests, c _p represents the standard deviation of CPU utilization of the p-th test round test cluster when receiving task requests, and m _p represents The standard deviation of the memory utilization rate of the p-th test round test cluster when receiving task requests, w _p represents the standard deviation of the network inflow bandwidth utilization rate of the p-th test round test cluster when receiving task requests, o _p represents the p-th test round The standard deviation of the network inflow bandwidth utilization rate of the test cluster when receiving task requests in each test round, the smaller the _dp , the better the load balancing performance of the test cluster;

(7d) The analysis and evaluation module calculates the total time block invalid request rate of each test round test cluster when receiving task requests, and the ratio of the number of overtime task requests per second and the total number of requests in the basic performance data of the test cluster when receiving task requests Ratio, as the invalid request rate per second of the test round; calculate the arithmetic mean of all invalid request rates within the duration of each test round, as the invalid request rate of the total time block of the test cluster in the test round, the total time block The smaller the invalid request rate, the better the load balancing performance of the test cluster;

(7e) The analysis and evaluation module selects the performance test index corresponding to the remaining test rounds according to the weight test rounds in the configuration file, and determines the weight of each index in the performance test index by the entropy method; use the largest test round The number of test rounds minus the weight value is used as the number of remaining test rounds, and the arithmetic mean weighted summation is performed on the performance test indicators corresponding to the remaining number of test rounds to obtain the load balancing performance score of the virtualization platform;

(8) Generate a test report:

The analysis and evaluation module generates a test report for the load balancing function test of the virtualization platform on the front-end virtualization platform monitoring page.

Compared with the prior art, the present invention has the following advantages:

First, because the data storage module in the device of the present invention can persistently store basic test data and support viewing of historical test data, the operation of the device is friendly to testers, which overcomes the need for testers in the prior art to test the data that has been tested for each test. The problem of repeated testing of the virtualization platform makes the device of the present invention have the advantages of reducing the complicated and redundant repeated testing work of testers and improving the efficiency of the load balancing function test of the virtualization platform.

Second, since the method of the present invention selects performance loss, availability, resource balance, and total time block invalid request rate as performance test indicators from the perspective of virtualization platform service providers and virtualization service buyers, and clearly proposes load balancing The performance formula, quantitatively and comprehensively evaluates the load balancing function of the virtualized service platform, overcomes the problem that the evaluation index in the prior art is too simple and difficult to quantify, and makes the method of the present invention have the practicability of quantitatively comparing the load balancing functions of different virtualized platforms The advantages.

Description of drawings

Fig. 1 is the block diagram of device of the present invention;

Figure 2 is a flow chart of the method of the present invention.

Detailed ways

The present invention will be described in further detail below in conjunction with the accompanying drawings.

Referring to accompanying drawing 1, the device of the present invention is clearly and completely described.

The device of the invention includes an environment configuration module, a data collection module, a data storage module, a load sending module and an analysis and evaluation module.

The environment configuration module is used to read the test configuration file; by interacting with the authentication service of the virtualization platform, obtain the management authority of the virtualization platform, and then interact with the virtualization platform through the authority to create a load balancer; according to In the configuration file, the test cluster virtual machine scale, the number of test cluster virtual machines, the test cluster virtual machine image, call the interface provided by the virtualization platform, create multiple virtual machines as a test cluster, and add the test cluster IP address to the load balancer in the load balancing pool.

The data acquisition module is used to test the benchmark performance data of the test cluster. When the task request is not sent, the basic performance data of the test cluster is collected by the virtualization platform monitoring system as the benchmark performance data of the load balancing function; in the load balancer When receiving a task request, the basic performance data of the test cluster is collected through the monitoring system of the virtualization platform.

The data storage module is used to store the measured benchmark performance data in a database for persistent storage.

The load sending module is used to generate a request of HTTP GET type according to the total number of task requests in the configuration file, the type of task requests, and the percentage of different types of task requests in the total number of task requests, through the preset task request type and url address One-to-one mapping relationship, modify the url address in the HTTP GET request, continue to build different types of task requests, and send the built task requests to the load balancer; count the number of each type of task requests that have been built, If the percentage of the number of task requests of this type to the total number of task requests in the configuration file satisfies the percentage of the number of task requests of this type in the total number of task requests in the configuration file, stop building this type of request; determine whether the number of load sending rounds is equal to the preset in the configuration file If it is the maximum number of test rounds, enter the next step to calculate the load balancing performance; otherwise, add 1 to the value of the number of load sending rounds and repeat the load sending step.

The analysis and evaluation module is used to calculate the performance loss of each test round test cluster when receiving a task request; calculate the responsiveness of each test round test cluster when receiving a task request; calculate each test round test cluster The degree of resource balance when receiving task requests; calculate the invalid request rate of the total time block of the test cluster for each test round when receiving task requests; according to the number of weight test rounds in the configuration file, select the number of weight test rounds For the corresponding performance test index, the weight of each index in the performance test index is determined by the entropy method; the maximum number of test rounds minus the number of weight test rounds is used as the remaining number of test rounds, and the performance corresponding to the remaining number of test rounds The test indicators are weighted and summed by the arithmetic mean value to obtain the load balancing performance score of the virtualization platform; the test report of the load balancing function test of the virtualization platform is generated on the front-end virtualization platform monitoring page.

The method of the present invention is described in further detail with reference to accompanying drawing 2.

Step 1, the environment configuration module reads the test configuration file.

Step 2, initialize the test environment.

The environment configuration module obtains the management authority of the virtualization platform by interacting with the authentication service of the virtualization platform, and then interacts with the virtualization platform through the authority to create a load balancer.

The environment configuration module calls the interface provided by the virtualization platform according to the test cluster virtual machine scale, the number of test cluster virtual machines, and the test cluster virtual machine image in the configuration file, creates multiple virtual machines as the test cluster, and adds the test cluster IP address to to the load balancing pool of the load balancer.

Step 3, collecting benchmark performance data of the storage test cluster.

The data acquisition module tests the benchmark performance data of the test cluster. When no task request is sent, the basic performance data of the test cluster is collected through the virtualization platform monitoring system as the benchmark performance data of the load balancing function. The data storage module stores the measured benchmark performance The data is stored in the database for persistent storage. The basic performance data includes CPU utilization, memory utilization, network inflow bandwidth utilization, network outflow bandwidth utilization, packet loss task requests, total task requests, timeout task requests number.

Since no task request is sent, the performance of the virtualization platform must be high at this time, which is close to the performance level of the virtualization platform under ideal conditions. The performance of the virtualization platform obtained in this way can be regarded as the benchmark performance of the virtualization platform.

Step 4, send a task request to the load balancer.

The load sending module generates HTTP GET type requests according to the total number of task requests, task request types, and the percentage of different types of task requests in the total number of task requests in the configuration file, and uses the one-to-one mapping relationship between preset task request types and url addresses , modify the url address in the HTTP GET request, continuously build different types of task requests, the task request types include computing-intensive task requests, IO-intensive task requests, and memory-intensive task requests, and send the constructed task requests to Load balancer; count the number of each type of task requests that have been constructed separately. If the percentage of the number of task requests of this type in the total number of task requests in the configuration file meets the percentage of task requests of this type in the total number of task requests in the configuration file, then stop building that type of request.

Step 5, collecting basic performance data of the storage test cluster.

The data acquisition module collects the basic performance data of the test cluster through the virtualization platform monitoring system when the load balancer receives the task request; the data storage module stores the measured basic performance data in the database for persistent storage.

Step 6, the load sending module judges whether the number of load sending rounds is equal to the maximum number of test rounds preset in the configuration file, if so, execute step 7, otherwise, add 1 to the value of the load sending rounds and execute step 4.

Step 7, calculate load balancing performance.

Due to the existence of the virtualization platform, the server resources of the virtualization platform can be allocated and managed more efficiently, which enables the virtualization platform to have the ability of rapid scaling, and the buyer of the virtualization platform service can apply for server resources of any size at any time , the load balancing function is used to maintain the stability of the application performance of the virtualization platform service buyer during the rapid scaling of server resources. There are two types of roles related to the load balancing function of the virtualization platform, namely the virtualization platform service provider and the virtualization platform service buyer. From the perspective of the virtualization platform service provider, the virtualization platform service provider will pay more attention to its external The performance loss and resource balance of providing services; from the perspective of virtualization platform service buyers, virtualization platform service buyers will pay more attention to the responsiveness of their purchased services and the invalid request rate of the total time block. To sum up, the present invention proposes four performance test indicators for indirectly reflecting the load balancing function of the virtualization platform, including performance loss, responsiveness, resource balance degree and total time block invalid request rate.

According to the following formula, the analysis and evaluation module calculates the performance loss of each test round when the test cluster receives task requests:

Among them, l _p represents the performance loss of the test cluster in the pth test round when receiving the task request, n _p represents the duration of the task request received by the test cluster in the pth test round, and i represents the test cluster in the pth test round Receive the serial number in the time series of the task request duration, b represents the benchmark performance of the test cluster, and the value of the benchmark performance is the CPU utilization rate, memory utilization rate, network inflow bandwidth utilization rate, and network outflow bandwidth in the benchmark performance data The average value of the utilization rate, q _pi represents the actual performance of the test cluster at the i-th second of the p-th test round, and the value of the actual performance is the CPU utilization at the i-th second of the p-th test round in the basic test data, The average value of memory utilization, network inflow bandwidth utilization, and network outflow bandwidth utilization. The smaller the value of _lp , the better the load balancing performance of the virtualization platform.

According to the following formula, the analysis and evaluation module calculates the responsiveness of each test round when the test cluster receives task requests:

Among them, r _p represents the responsiveness of the p-th test round test cluster when receiving task requests, e represents a natural constant, and D _p represents the number of task requests that the p-th test round test cluster loses packets when receiving task requests, Q represents the total number of task requests constructed by the load sending module, the smaller the value of r _p , the better the load balancing performance of the virtualization platform.

According to the following formula, the analysis and evaluation module calculates the resource balance of each test round when the test cluster receives task requests:

Among them, d _p represents the resource distribution balance degree of the p-th test round test cluster when receiving task requests, c _p represents the standard deviation of CPU utilization of the p-th test round test cluster when receiving task requests, and m _p represents The standard deviation of the memory utilization rate of the p-th test round test cluster when receiving task requests, w _p represents the standard deviation of the network inflow bandwidth utilization rate of the p-th test round test cluster when receiving task requests, o _p represents the p-th test round The standard deviation of the network inflow bandwidth utilization rate of the test cluster when receiving task requests for each test round, the smaller the _dp , the better the load balancing performance of the test cluster.

The analysis and evaluation module calculates the total time block invalid request rate of each test round test cluster when receiving task requests, and the ratio of the number of overtime task requests per second to the total number of requests in the basic performance data of the test cluster when receiving task requests, as The invalid request rate per second of the test round; calculate the arithmetic mean of all invalid request rates within the duration of each test round as the total time block invalid request rate of the test cluster in this test round, and the total time block invalid request rate The smaller the value, the better the load balancing performance of the test cluster.

The analysis and evaluation module selects the performance test indicators corresponding to the remaining test rounds according to the weight test rounds in the configuration file, and the performance test indicators include performance loss, responsiveness, resource distribution balance, and invalid request rate of total time blocks , the weight of each index in the performance test index is determined by the entropy method. The entropy method is an objective weighting method widely used in the field of comprehensive evaluation. Determining the weight of the index reduces the interference of human factors on the evaluation results; using the maximum number of test rounds minus the number of weight test rounds as the remaining number of test rounds, the arithmetic mean weighted calculation of the performance test indicators corresponding to the remaining number of test rounds is carried out. and to get the load balancing performance score of the virtualization platform.

Step 8, generate a test report.

The analysis and evaluation module generates a test report for the load balancing function test of the virtualization platform on the front-end virtualization platform monitoring page.

Claims (5)

1. A test method of load balancing function and performance of a virtualization platform is characterized in that a performance test index capable of indirectly reflecting the load balancing function is calculated by collecting basic performance data of a test cluster connected with a load balancer, and the weight of the performance test index is determined according to an entropy method; the method comprises the following specific steps:

(1) The environment configuration module reads the test configuration file;

(2) Initializing a test environment:

(2a) The environment configuration module acquires the management authority of the virtualization platform through the authentication service interaction with the virtualization platform, and then interacts with the virtualization platform through the authority to create a load balancer;

(2b) The environment configuration module calls an interface provided by the virtualization platform according to the test cluster virtual machine scale, the test cluster virtual machine number and the test cluster virtual machine mirror image in the configuration file, creates a plurality of virtual machines as a test cluster, and adds the test cluster IP address into a load balancing pool of the load balancer;

(3) Collecting and storing benchmark performance data of the test cluster:

the data acquisition module performs a benchmark performance data test on the test cluster, when a task request is not sent, basic performance data of the test cluster are acquired through the virtualization platform monitoring system and serve as benchmark performance data of a load balancing function, and the data storage module stores the measured benchmark performance data into a database for persistent storage;

(4) Sending a task request to a load balancer:

the load sending module generates a request of an HTTP GET type according to the total number of task requests, the types of the task requests and the percentage of different types of task requests in the total number of the task requests in the configuration file, modifies url addresses in the HTTP GET requests through a preset one-to-one mapping relation between the types of the task requests and the url addresses, continuously constructs the task requests of different types, and sends the constructed task requests to a load balancer; counting the quantity of each type of constructed task request respectively, and stopping constructing the type of request if the percentage of the quantity of the type of task requests in the total quantity of the task requests in the configuration file meets the requirement that the quantity of the type of task requests in the configuration file in the total quantity of the task requests;

(5) Collecting and storing basic performance data of the test cluster:

the data acquisition module acquires basic performance data of the test cluster through the virtualization platform monitoring system when the load balancer receives the task request; the data storage module stores the measured basic performance data into a database for persistent storage;

(6) The load sending module judges whether the load sending round number is equal to the maximum testing round number preset in the configuration file, if so, the step (7) is executed, otherwise, the step (4) is executed after the value of the load sending round number is added with 1;

(7) Calculating the load balancing performance:

(7a) According to the following formula, the analysis and evaluation module calculates the performance loss of each test turn test cluster when receiving the task request:

wherein l _p Representing the performance loss of the p test round test cluster at the time of receiving the task request, n _p Indicating the duration of the p test round test cluster receiving the task request, and i indicating the p test round test cluster receiving the task requestThe serial number in the time sequence of the duration, b represents the reference performance of the test cluster, the value of the reference performance is the average value of the CPU utilization rate, the memory utilization rate, the network inflow bandwidth utilization rate and the network outflow bandwidth utilization rate in the reference performance data, q _pi The practical performance of the test cluster at the ith test turn in the ith second is represented, and the value of the practical performance is the average value of the CPU utilization rate, the memory utilization rate, the network inflow bandwidth utilization rate and the network outflow bandwidth utilization rate at the ith test turn in the basic test data, l _p The smaller the value of (A), the better the load balancing performance of the virtualization platform;

(7b) The analysis and evaluation module calculates the responsiveness of each test round test cluster when receiving the task request according to the following formula:

wherein r is _p Representing the responsiveness of the test cluster of the p test round when receiving the task request, e representing a natural constant, D _p The number of task requests representing packet loss when the p test round test cluster receives the task requests, Q represents the total number of the task requests constructed by the load sending module, and r represents the total number of the task requests _p The smaller the value of (A), the better the load balancing performance of the virtualization platform;

(7c) According to the following formula, the analysis and evaluation module calculates the resource balance degree of each test round test cluster when receiving the task request:

wherein d is _p Representing the resource distribution balance of the p test round test cluster when receiving the task request, c _p Represents the standard deviation of CPU utilization rate, m, of the p test round test cluster when receiving the task request _p Represents the standard deviation of the memory utilization rate, w, of the p test round test cluster when receiving the task request _p Denotes the p testNetwork inflow bandwidth utilization standard deviation, o, of turn test cluster when receiving task request _p Represents the standard deviation of the network inflow bandwidth utilization rate of the p test round test cluster when receiving the task request, d _p The smaller the cluster load balancing performance is, the better the cluster load balancing performance is tested;

(7d) The analysis and evaluation module calculates the invalid request rate of the total time block of each test round test cluster when receiving the task request, and the ratio of the overtime task request number per second to the total request number per second in the basic performance data of the test cluster when receiving the task request is used as the invalid request rate per second of the test round; calculating an arithmetic mean value of all invalid request rates within the duration of each test round, wherein the arithmetic mean value is used as the invalid request rate of the total time block of the test cluster of the test round, and the smaller the invalid request rate of the total time block is, the better the load balancing performance of the test cluster is;

(7e) The analysis evaluation module selects performance test indexes corresponding to the weight test rounds according to the weight test rounds in the configuration file, and determines the weight of each index in the performance test indexes through an entropy method; subtracting the weight number of testing rounds from the maximum number of testing rounds to serve as the number of remaining testing rounds, and carrying out arithmetic mean weighted summation on performance testing indexes corresponding to the number of remaining testing rounds to obtain a load balancing performance score of the virtualization platform;

(8) And (3) generating a test report:

and the analysis and evaluation module generates a test report of the load balancing function test of the virtualization platform on the front-end virtualization platform monitoring page.

2. The method according to claim 1, wherein the basic performance data in step (3) includes a CPU utilization rate, a memory utilization rate, a network inflow bandwidth utilization rate, a network outflow bandwidth utilization rate, a packet loss task request count, a task request total count, and an timeout task request count.

3. The method for testing the load balancing function and performance of the virtualization platform as claimed in claim 1, wherein the task request types in step (4) include a compute intensive task request, an IO intensive task request, and a memory intensive task request.

4. The method for testing the load balancing function and performance of the virtualization platform according to claim 1, wherein the performance test indicators in step (7 e) include performance loss, responsiveness, resource distribution balance, and total time block invalid request rate.

5. The method for testing load balancing function and performance of a virtualization platform according to claim 1, wherein the entropy method in step (7 e) is an objective weighting method widely used in the field of comprehensive evaluation, and determines the index weight according to the size of information provided by each index observation.

Images