After the performance test , One of the most important tasks may be the analysis of performance problems . How to do performance analysis , It may be a difficult problem for novices . For starters , The idea of analyzing performance problems may be as follows , It is summarized in four words “ More flowers ”.
This way of thinking , We can understand it as divergent mode , While brainstorming , It is useful to , But the analysis of performance problems after doing things , It doesn't seem so handy . So for this type of post analysis , You can still sum up some rules , Then the purpose is to save unnecessary trial and error time .
According to the author's experience , Before you really start analyzing the problem , First of all, we need to ensure that our performance test is effective . This is the key point , If you analyze the problem for half a day and find that the test is invalid , That could be a sad story . Then maybe a friend wants to ask , What is an invalid performance test ？ It may include but is not limited to the following situations ：
- The test scenario was not communicated clearly with the requirements team or the product team , The result of the test is not what the team needs . Therefore, at this time, it is necessary to communicate the needs with relevant personnel , Then refine it into performance requirement points , The last and most important thing is to work with the product team （ The product manager ） Identify the needs , Is there a deviation in your understanding .
- There is a problem with the test script , Sometimes time may be tight , The performance test engineer didn't check the script well , In fact, there is a problem with the script , For example, without a request , Or lack of key breakpoint checking, etc . So in order to deal with these problems , Before the official run , Some verification work needs to be done . such as , The official run is 100 Concurrent run 20 minute , Then in the verification phase, you can 20 Concurrent run 5 minute , The purpose is to eliminate script problems .
- During the test run , The system under test has changed , For example, deployment is in progress , So the test results after running are not referential .
To sum up, the above situation is as follows ：
Before analyzing the problem , These ineffectiveness problems deserve our attention . secondly , If you want more test results , Provide reference for our analysis , We need to consider more monitoring means . This requires good cooperation with other colleagues , such as DevOps Or implementation engineer , Try to have a comprehensive monitoring of the tested system . such as Datadog perhaps Dynamics And so on APM Tools , You can also develop some script programs to monitor some specific service status .
So in performance analysis activities , Need better teamwork ( communicate ), To find the problem as soon as possible , solve the problem .
The next article will continue “ How to do performance problem analysis ” series . Please pay attention , thank you ！