Advanced performance test series 4. premise, tool and process of performance test

Catalog

• One 、 The premise of performance testing
  • 1、 Research on the necessity of performance testing --- Key item assessment
  • 2、 Testability --- It can be quantified as performance index value
  • 3、 Insufficient corporate servers , Before the project goes online , Can I use the production environment for performance testing first ？
  • 4、 Some of the company's projects have been launched but no users have used them , such as 3 The monthly project has been completed ,4 It is only available to users in months . In this case , Can I use the production environment for performance testing ？
• Two 、 Performance testing tools
• 3、 ... and 、 Performance testing process

One 、 The premise of performance testing

1、 Research on the necessity of performance testing --- Key item assessment

Do a performance test , The first thing to do is Evaluation of key items .

This product of the company needs to be tested . You need to know which functions need to be tested first , Which functions will be tested for performance .

If the product has 100 Features , There is not even a priority , I took the 100 The three functions are tested together , So I lost the watermelon and picked up the sesame , Nothing .

1） Department in charge 、 Regulatory review ;

2） Involving the safety of life and property ;

Like the banking system , It's been revised recently , The new version replaces the old one , If there is no performance test , In case of high concurrency , data （ money ） There is a problem , It's going to be a lot of trouble .

3） Big new systems ;

When the new version replaces the old version , It must be the performance test of the new version , Otherwise, the response time will be longer , Worse performance , The worse the user experience .

4） The core system ;

For example, there are 100 One business , It must be doing Core business , Or the most used by users first .

The priority and importance of the business need to be assessed , There is not so much time for performance testing of any business .

5） Structural adjustment ;

If the company develops projects with 1.8 Of jdk, Now there is jdk1.11, Find out jdk1.11 There are many new features .

I just put jdk Upgrade to 1.11 edition . After upgrading , There is no problem with the function test , But there may be performance problems .

jdk This thing is the bottom thing , To run java The code must have jre Operating environment , Installed in the operating environment jdk、jre.

This is the underlying dependency , If the performance of this underlying dependency does not meet the requirements . After the product goes online , The function is OK , But once the user comes up, there will be performance problems .

So this bottom adjustment , It also requires comprehensive performance testing .

6） There's a surge in business ;

This kind of promotion , Performance testing must be done before launching .

Although this is not the core business , This kind of promotion is over , You may not use this page again next time , But because of its particularity , Performance tests are also required .

7） Major defect repair .

There are some old pits in the projects of some companies , Finally, it took time and cost , Changed a lot of things , One day he repaired the pit .

After the pit is repaired , It's a big influence , The functional tester shall work overtime to test , Performance testers also have to work overtime to test performance .

Because when the old pit was repaired , Changed the bottom things , The performance of the underlying things has not kept up , No problem with the function , But there may be a lot of code that needs to be tweaked .

Repair of major defects , Generally, the bottom level will be changed widely , After the repair, performance test is required .

2、 Testability --- It can be quantified as performance index value

If your company's project is product oriented , The product personnel put forward the demand for this performance , He doesn't know about performance testing .

If your company is this kind of project oriented , The project manager does not understand performance testing .

So what they need , It's hard to do performance tests . But you also have to do performance tests , Then you need to master the knowledge and skills of performance testing , Come and talk to your supervisor Communicate repeatedly to determine performance indicators .

3、 Insufficient corporate servers , Before the project goes online , Can I use the production environment for performance testing first ？

The project has not been launched yet , The code has not been put into the production environment , Performance testing is not possible .

Although some enterprises switch the production environment to the test environment , The key is that the code has not been put into the production environment , Since the code has not been updated in the past , You can't do performance tests .

If you don't finish testing, put the code into the production environment , Let you do a performance test , This is equivalent to being online .

If your company has a grayscale publishing environment , This is OK .

If the production environment has 30 Taiwan machine , You can split several machines from them and update the code first , Use some real users to experience this new function .

See if the new features meet their requirements . If meet , Then update the remaining machines with new codes .

This actually divides the production environment into 2 An environment , One is old code , One is the new code .

You can use this environment , But pay attention to the problem of database . When the production environment is a grayscale publishing environment , Generally, the database will also be used for 2 individual .

Especially if the database has a table structure change , If you don't have two databases , Grayscale publishing environment is not available .

But you have to switch the database of the grayscale publishing environment , You cannot publish the database of the environment in grayscale , Because there will be dirty data in the gray-scale publishing environment after use .

Because the grayscale publishing environment , One day it will be transformed into a formal environment .

The dirty data is in the database of the gray-scale publishing environment , One day, it will become dirty data in the production environment .

4、 Some of the company's projects have been launched but no users have used them , such as 3 The monthly project has been completed ,4 It is only available to users in months . In this case , Can I use the production environment for performance testing ？

The product has been released online , But there are no users . Or it will iteratively release new functions to the production environment , Let potential customers experience first , After experiencing , Meet the customer's requirements , There will be real paying users to use .

This situation , If you do a performance test, there will also be dirty data . Cleaning up dirty data in the production environment is not clean .

Backup first , Then do performance tests , After that, use the backup to restore , The success rate of cleaning up dirty data in this way is not high .

And there are cost and operational risks , Under normal circumstances, the company will not let you do this .

Two 、 Performance testing tools

Open source ：jmeter

java Development 、 Cross platform 、 Quick version update （ Suggest v5.1.1 in the future +jdk1.8）.

business ：loadrunner

Performance test benchmarking software 、c Language 、 Domestic cracking （< lr11)、lr12 Free trial 50 Limit the number of users 、 Slow update .

Since the research ：python+locust

python The language is developed by itself .

3、 ... and 、 Performance testing process

1. Preparation for performance test

• 1） Demand analysis ----- Familiar with the business .
• 2） Clear performance test objectives （ Index value ）.
• 3） Learn about software functions 、 framework .
• 4） Specify the test plan , Do a good job of workload assessment .
• 5） Develop a test model （ Edit test cases ）.

It is different from functional test , The same thing is that we all need to do demand analysis .

Functional testing focuses on a single person initiating a request , The purpose is to find bug. But professional performance testers focus on multiple people making requests , At this time, the function appears bug, I don't care .

Do not directly understand the business , But understand the overall data flow 、 The overall structure 、 Overall function .

Implementation of functional logic of specific interface , It doesn't matter if I don't know . However, it is necessary to understand how the functions interact with each other as a whole , Which function needs to depend on what data .

Server architecture , How to exchange data between services , What kind of configuration , This has to be clear . Because we need to know these things before setting up the environment .

Workload assessment ： It will be better than functional test 、 Automated testing takes a long time . Equal work , The time consumed is probably theirs 2-3 Times of time .

Not all functions need to be tested , Only some functions are tested .

Test cases are transformed into performance test scenarios 、 Load test scenarios, etc .

2. Build a performance test environment

• 1） Tool selection and preparation .

Different agreements , Will use different tools .

• 2） Build the environment of the system under test （ The server 、 Service version update 、 Database data preparation 、 Monitoring the environment ）.

Before testing , The data level of the database should be prepared .

• 3） The network configuration .

3. Performance test script development

• Select protocol .
• Making scripts .
• Debug script .
• Validation script .

You can make scripts in the test environment , Debug script .

In development , Functional testers are also testing , The scripts written by performance testers can directly connect to the test environment .

In the future, only the environmental ip Connect to the performance test environment , You can perform performance tests in the performance test environment .

Script development is not affected , Because the code of the functional test environment is updated before the code of the performance test environment .

Debug scripts and validation scripts , It will be gradually transferred to the performance environment .

4. Performance test script execution

• A test run .
• Scene execution .

Use test cases in scripts , Perform scene conversion , Execution scenarios .

To build a monitoring environment , Collect performance data for test results . It is used to analyze and tune the performance test results later .

5. Result analysis and tuning

• 1） The analysis is based on ： The result chart .
• 2） Analysis methods ： Server hardware bottleneck > Network bottleneck > The server os bottleneck （ Parameter configuration 、 database 、web The server ）> Application bottleneck （sql sentence 、 Database design 、 Business logic 、 Algorithm ）.
• 3） tuning .
• 4） Modify the script or scene .

6. Test report and result tracking

• Performance test report .
• Performance test problem tracking .

The performance problems found cannot be solved this time . For example, a memory problem is found , It may be the structural adjustment of the code to solve this problem .

It can't be solved in the current period .

Then it will be solved later , And then repeatedly verify and debug , See if this problem has been really solved .