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

Now that we understand the performance test load model we mentioned earlier, how can we test based on this model?

Let's talk about the performance testing strategy based on the load model we introduced earlier.

The performance testing strategy is to select different test types according to different test purposes and different concerns.

Review the formula of the test model that we described earlier, MirrF {L _ Magneo _ T} and the definition of the scenario that is the profile of the system running. The so-called test type refers to selecting different profiles to run different periods of time. As shown in the following figure:

In the three-dimensional model we put forward in the model chapter, we put forward such an assumption: with the change of time, the number of users increases uniformly.

In this assumption, the model of the system is shown in the figure above, in which we take three profiles, S1, S2 and S3.

S1 profile means that from zero, with the increase of the number of users, the coincidence degree of the scene increases gradually, and the resource reuse rate of the cache increases, so as the number of users increases, the average performance of the system increases faster.

S2 profile means that with the increase of the number of users, the coincidence degree of the scene has increased to the limit, and the resource consumption caused by the continued increase of users begins to increase linearly with the increase of users. the average performance of the system slows down with the increase of the number of users, but the number of users that the system can support can continue to increase.

S3 profile is to continue to increase the number of users, the capacity of the system has increased to the limit, at this time, continue to increase users, at this time began to appear resource requisition, began to appear frequent waiting, the average new energy performance of the system began to decline with the increase of users, and the capacity of the system began to reach the limit and no longer increased.

Note: the average performance of the system refers to the "increase" rate of performance indicators (such as response time)

-shame-shame-of-points-cut-line-

Policy selection load test

Load testing refers to focusing on the load attribute in the performance test model, that is, the variation is L. O and T remain the same.

Corresponding to the above three profiles, that is, the load test runs for the same period of time under the three scenarios of S1, S2 and S3 respectively, to evaluate the performance of the system and the performance change curve of the system under different loads. (or there are more scenarios, but at least include S1 scenario as the initial test target, S2 scenario as the system expected target, and S3 scenario as the test termination target)

Specific scenario settings such as S1 scene run for half an hour, then the load is increased to the load of S2 section, continue to run for half an hour in this scene, then continue to increase the load to the load of S3 section, and run again for half an hour in this scene.

Based on this, it is generally necessary to confirm the capacity of the system (S3 profile), confirm the best performance of the system (S2 profile) and other testing purposes, or load testing is generally required when the TPS is expected to reach 100, or the system supports 500 users to use the system online.

Pressure testing

Stress test means that the load is the maximum, that is, the attribute L remains the same at this time, O remains the same, and T changes, that is, the running time changes.

Corresponding to the three profiles above, it means that the system runs under the load of section S3. When T is less than or equal to 2, it is what we often call stress test; when T is greater than or equal to 8, it is what we often call load test.

The stress test is mainly concerned with the stability of the system, that is, whether the system can run smoothly for a period of time under the possible maximum load without exception.

For the test purpose with the requirement of stability, it is necessary to choose the stress test when determining the test strategy.

Concurrent testing

Concurrent testing, that is, the object concerned is the load object, that is, O is the amount of change in MSecretF {LMagol O T}, and each round of testing only pays attention to the performance of a certain test object under a certain load.

In a more rigorous process of concurrent testing, multiple rounds must be carried out, that is, the object O of each round of testing changes from its original node to its direct subordinate node, in order to locate the problem at a deeper level.

From this, it also leads to the keywords that we choose to carry out concurrent testing to locate the deeper problems. Such as resource contention, resource leakage and so on. Generally, for the purpose of testing, you need to test a function or component, and you need to choose concurrent testing.

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.