How to write a performance test plan

Catalog

One ． brief introduction

Two ． Performance test requirements

3、 ... and ． Test environment

Four ． Data preparation

5、 ... and ． Testing tools

6、 ... and ． The test strategy

7、 ... and ． Manpower and time arrangement

---

One ． brief introduction

　　 There is no need to describe too much in the introduction , Nothing but the background of the project , The reason for this performance test , And the scope of performance test coverage, etc , Almost all project documents begin with a brief description of the project .

Two ． Performance test requirements

Look for the tested object and pressure point

　　 The object to be tested is not imagined out of thin air , But through analysis and systematic data collection . Take down several typical pressure points

Sign in ： For general systems , Login is the premise of user operating system , If the user can't log in at all , Then other functions will be useless . For example, online games , When opening a new suit , Players broke their heads just to log in .

Inquire about ： Query generally consumes system and database resources . The query function of search engine is typical , If you enter something in the input box , There will be no results for a long time . I want to be called “ Internet access ” Search engines will not exist .

transaction ： For some e-commerce systems , The performance requirements of the transaction process are very high , If the transaction process takes users a long time . I'd rather go shopping in the supermarket . Of course , In addition to the speed of trading , The requirement for the success rate of the transaction is also very high . Otherwise , The damage is also immeasurable .

The system under test should be the most important and basic function , It is also the most frequently used function by users .

                   

 

General performance requirements include ：

System capacity ： How many users can the system accommodate to register .

Number of visits ： Number of users accessing the system at the same time .

Concurrency number ： The number of concurrent operations performed simultaneously , There should be a combination of concurrent numbers of different operations in a system （ Generally, users with permission to operate ）.

The maximum number of users and the best number of users of the system : The maximum number of concurrent users the system is bearing , The concurrent user data that the system bears in the best state .

response time ： The time interval between the user submitting an operation and getting a response .

Throughput rate ： The system processes every second TPS

　　 A key factor in performance testing is stress , Performance is the performance under the maximum pressure that the system design meets . The concurrency number should not be less than the peak value of the normal operation of the system , The total amount of data is not less than the normal operation of the system 3 Monthly data volume .

　　 When describing the number of concurrent users , There will always be corresponding time limit . In essence, the performance index of the system should be described by the number of requests processed by the system in unit time and the request response time . The number of requests that can be processed per unit time is the business throughput of the system . The number of virtual concurrent users can be converted using the following formula ： （ Number of real users × Number of requests per real user ）/( Total request response time + Real users always think about time )=( Number of virtual users × Number of requests per user )/( Total request response time + Virtual users always think about time )= throughput .

3、 ... and ． Test environment

The test environment here mainly refers to the software and hardware environment and network environment .

　　 The author believes that the performance test is best carried out in an independent environment , In this way, it will not be disturbed by the outside world , The test data is independent and valid . If you now stress test a website that is already online , So the data you get is not independent , Because when you're doing a stress test , Other retail investors are also accessing the system .

Software environment ：

　　 The software environment here mainly refers to the environment in which the project runs , For example, what kind of operating system 、 middleware 、 And the database .

Hardware environment ：

　　 The hardware environment here mainly includes the internal components of the host ,cpu、 Memory 、 Disk and motherboard 、 Network card, etc , Transmission media and routers should also be taken into account ,

The network environment ：

　　 In addition to considering that the testing machine and the system server are carried out in a LAN in the network environment , We should also ensure the independence of this network . If during the performance test , Other machines are also consuming router resources . Then the router will also affect the transmission speed of the database .

Four ． Data preparation

　　 In many cases , We need to prepare test data , For example, the system does not allow repeated login of the same user , Then you must generate legal user data . Sometimes the system needs to be queried and tested , Only when the system has a certain amount of data can the real performance of the system be verified . There are two pieces of data and 20 million pieces of data in a database , Same query operation , The pressure on the system is completely different .

The analysis of data required by the system can refer to the following methods ：

　　 Historical data analysis is helpful to determine the data magnitude . Start with historical data , Find out the peak data volume .

　　 Start with other similar or identical systems , Data analysis , Find out the peak data volume .

　　 When there is no history or related system to refer to , It is necessary to estimate the performance data of the system , Including system capacity , Concurrency number and other data , After the estimation, it will be reviewed or revised by relevant personnel , Test according to the agreed performance index .

…………

　　 The test data should be the same as the real data , If you can get the real system running 3 Months of data , On this basis, we can conduct performance test .

　　 About data generation , We can wish a tool complete , Such as database data generation tools , Size file generation tools, etc .

                  

 

5、 ... and ． Testing tools

　　 I've already described how to analyze requirements , After the requirements are determined , We can consider what kind of tools to introduce to meet the performance requirements .

　　 Of course , When introducing tools, in addition to considering whether they can meet the requirements , The cost of tools should also be considered , It doesn't just mean the cost of purchasing tools , And the cost of learning tools for testers .

　　 About the choice of testing tools , There will be a separate chapter later , I won't go into details here .

　　 If the performance testing tool you choose is not powerful enough , You may also need other auxiliary tools . If jmeter utilize badboy To record the script , It can improve the efficiency of script development . Performance counters may also be required to record the performance of software and hardware during stress testing . Such as monitoring server cpu、 Memory counter , Monitoring tool for recording middleware logs , Monitoring tools for monitoring database performance, etc .

6、 ... and ． The test strategy

　　 For a specific business system , Users usually visit in different time periods of the day . In different periods of time , Users use business systems with different frequencies , And the system's busy degree is different . Under certain conditions , Users may access a business system in a short time . For example, for the document processing subsystem , There may be a situation where a large number of users view and handle an official document in a short time .  During the performance test , Should be used “ The principle of considering the worst case ”. That is, users should use the business system most frequently 、 Test the function of the system under the maximum pressure on the system , Judge whether each function and page can meet the performance requirements , Is the response time of the system too long .

　　 On the other hand , The performance of the system must be verified “ The coverage is comprehensive ”. Although the use frequency of each function in the system is different , Some functions are used less frequently than others , But when it comes to performance testing and optimization , These functions cannot be ignored , When compiling test cases, you can't just select the most commonly used functions . For example, all users who visit my list will be notified , But generally only 5% Users will use the system setting module to find the information of a user ; But in the test , We can't because the function of viewing user information is used relatively infrequently , And ignore the test of this function . therefore , During the system performance test here , For different businesses , The access proportion of users should be reasonably allocated .

　　 In terms of testing strategy , We should also consider , Performance of the same system in different hardware environments . So that the system can meet the requirements , The hardware configuration can also reach an optimal state . It is also a waste to add too much hardware to meet the demand . Besides, adding hardware equipment can not solve all performance problems .

7、 ... and ． Manpower and time arrangement

　　 The last item , According to the schedule and scale of the project , To arrange manpower and time . For large performance tests , Demand research in the early stage of the project , Deployment of the environment , Purchase or development of tools , Study and use of testing tools by personnel , After the performance test , Analysis and tuning of later data . All need personnel arrangement . Yes, you can need professional , System Engineer 、 Database Engineer 、 Software Development Engineer 、 The network engineer and Performance Test Engineer shall participate and cooperate to complete . Not a performance tester can do it all .

　　 I heard that , Best cattle x Performance testing , Performance testing teams from more than a dozen cities in several countries need to synchronize . The preparatory work will take several months . How to synchronize the control performance test . Summary and analysis of later test data . It's a very complicated process . This example needs to be verified , What I want to say is , For the performance test of large projects , People and timing are also crucial .