Shulou(Shulou.com)06/03 Report--

From the purpose of testing, performance testing can be divided into three categories: evaluation test, confirmation test and tuning test.

The main purpose of evaluation testing is to evaluate the performance of running under certain conditions. Such as testing the performance of the system in different environments, the performance of the software changes with the change of the number of users or the amount of data. Evaluation tests often do not clearly pass the goal, but give some conclusions of this evaluation according to the analysis and comparison of the test results.

The main purpose of confirmation test is to verify whether the system meets the specified performance requirements. Usually, the purpose of the test may be to test whether the average response time, or throughput, or the number of concurrent users of the system under certain conditions meet the specified requirements. The confirmation test must have a very clear passing standard, and the test results are compared with the established passing standards, and the conclusion of whether the system under test is confirmed or not is given.

The main purpose of tuning testing is to find out the performance bottleneck of the system under test, analyze the causes of system performance defects, and optimize the performance of the system to improve the performance of the software. For example, the performance of the software is tested to determine whether there are memory leaks, thread leaks, unreasonable contention for resources or deadlocks and other problems, and optimize its performance. Tuning testing is generally a concurrent test of a single business scenario, so that the problem can be more specifically simulated and located more quickly.

According to different test types, it is necessary to design the scheduling of load objects and the running process such as thinking time and iteration time in the process of test execution, in order to be closer to the requirements of different test types. For these different designs, three different models are proposed here for your reference.

Waterfall model

The waterfall model is to add a test object to provide load at regular intervals according to the total length of the test. Through this combination model, we can ensure that every time period, the current load model is different, not only the load has changed, but also the load object has changed at the same time.

This model is mainly used in confirmation testing, which can not only confirm whether there is a problem in the system, but also quickly confirm which load object leads to the performance problem according to the point of time when the problem occurs. This model is more meaningful for the positioning of the problem.

The disadvantage of this model is that according to the point in time when the problem occurs, we can only confirm which kind of load object is the cause of the performance problem. However, it does not answer the performance problem caused by the load object and which load object provides the load at the same time. For example, the following figure shows that the system began to experience problems after CASE3 was added to the load. We have no way to know whether the specific problem is caused by the simultaneous operation of CASE3 and CASE1. It is also because CASE3 and CASE2 are running at the same time that cause system problems.

Fishbone model

Based on the shortcomings of the waterfall model mentioned above, we propose another model, the fishbone model. The fishbone model covers all possible combinations through various combinations of different load objects.

As shown in the following figure: from the combination of pairwise and trinity of different load objects to all the load objects running at the same time. But the disadvantage of this model is that when a scene contains a large number of load objects, this possible combination increases exponentially. Therefore, scenarios with more than 4 load objects are not suitable for this model.

The model is also mainly used for confirmation testing. For the case of less load objects, we can directly use this model to design the load object scheduling; for the case of more load objects, it should be combined with the waterfall model, first through the waterfall model to design and locate the load objects that cause system problems, and then narrow the scope and then accurately locate through the fishbone model.

Hybrid model

In the actual testing process, many times, we do not have much time to test all possible combinations of load objects, even if we first narrow the scope through the waterfall model, it still takes a long time, we tend to run all load objects at the same time, in order to be able to cover as many load object combinations as possible in a fixed time We often need to set the thinking time and iteration time to a random range. The maximum value of the time range for each test object is recommended to be 1 to 7 or 1 to 13 of the time it takes for the load to execute once, to minimize the coincidence of combinations caused by meeting all common multiples.

This model is the model we use most in daily testing. the disadvantage of this model is that even if there is a problem with the test, it cannot be caused by which load objects are running at the same time. Therefore, most of the time, the model is mainly suitable for evaluation testing.

So far, this is the end of ten articles on the performance testing load model. I will gradually talk about some ideas for the construction of the performance testing system when I have time.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.