Here, our performance tests are divided into performance indicators, testing and performance stress tests.
The performance index test is to determine whether the product can meet the user's needs by simulating the real concurrent user operation; For example, wechat needs to support the access of 500000000 users as a real user needs, but there is no way we can actually access 500000000 users. Only through the simulation of 500000000 real users to access, if we achieve this goal through the simulation, we are satisfied with this demand.
The performance pressure test is to determine whether the function of the product is still normal and whether the system itself is normal. For example, when the CPU and memory take up more than 80%, the system crashed.
1, performance index test:
First, we need to know if the corresponding demand for performance indicators is reasonable and clear enough. For example, we need to support 30,000 users and pay online at the same time. Then we'll have the following questions:
1.How does 1, 30,000 users come and infer from what kind of model is reasonable?
2.Is the method of payment determined, whether or not the way of payment is affected?
3, the process of payment (generally should not be non-atomic operations) should be all consistent?
4.Is it necessary to choose to be successful, or do you need a certain number of users to choose to pay for the failure?
When we really understand the demand, we can better stand on the user's angle to think about our performance index test process. Also can let our user index test process more consistent with the real user uses the scene.
After the analysis of the demand, we can begin to design the corresponding performance index for the requirements, and we need to analyze the specific needs and whether there is a test method to cover it. For example, how are we going to simulate the users of the 3000W? This time, we may need to go to our own development test tools, or need to develop to provide us with the corresponding interface.
After the research and determination of the test method, we began to cover the test according to the specific test method. Of course, the process of testing is generally not smooth. I t is also necessary for us to have a certain analytical ability, for example: when we find out and send the number of users to 10 w, we are not going to go. Is it possible for us to check out the possible reasons? Tools themselves, network problems, interface problems, system load problems, and so on.
According to our established test plan, after completing the performance index test, we need to analyze and organize the whole performance test process and results. This allows us to submit a high-quality performance test report.
2, performance pressure test.
The performance pressure test is the same as the function test, and it is mainly a process of evaluating the quality by finding bugs. So here are two questions:
1. What kind of problem is a performance problem (that is, the difference between functional problems and performance problems)?
2. How to ensure the quality of performance (that is to find all performance problems)?
Generally speaking, when the system is in a certain load, the problem is a performance problem. For example: with QQ, you may have a QQ user to log on when the server is no problem, but 10000 QQ while logging on and sending data when the server may have a problem. If there is a big difference between the resources that are occupied, the logical problems that can be found through the test function are not a performance problem. Then, how to ensure the quality of performance? This is a way to ensure that the quality of performance is guaranteed by covering the performance of the performance business logic. As follows:
1. the business logic is very detailed.
2. The principle of extracting performance points based on business logic is to determine whether a single connection and multiple connections will affect the logic, as long as some of them should be extracted. Such as the application and release of resources, multi-thread working way.
3, according to the performance of the performance of the performance of the writing, we must ensure that these performance test methods can achieve the purpose of testing.
4, analyze whether some test methods or checkpoints can be realized through script or automation, to save test time.
5, learning corresponding test tools, the knowledge of tools can effectively improve their test efficiency.
6.Like the function test, the test process must be analyzed, and it is best to analyze every performance bug, and see if there is a place that may not be considered. Make supplementary tests and perfect use cases (and analyze why there is no consideration at the time) to form experience documents.
7, strengthen the study of the measurement system, because the performance pressure test itself is a whole, only to be aware of the systematic understanding of the depth to be able to analyze the more accurate.
8, try to check and locate the problem, know where the whole performance bottleneck is, and know how to cover it.