Performance testing steps

To: gotesting software testing Alliance Forum http://www.78test.com

Recently, during the training for new employees, we summarized and sorted out the performance testing steps and put them here for everyone to make a decision...

Collection of performance testing requirements: This step starts from the ground up, from the endless process, to the collection of performance indicators in product requirements. We can generally try to use the hardware environment on which the software depends for the purpose of performance testing, in terms of software architecture, if you have professional product personnel, you need to worry about it. If you have non-professional product personnel, you have to work hard. The work at this stage determines the success or failure of the later design. The specific method will be used to write another article after I finish the summary.

Access Model Modeling: the number of users and usage habits of different products are different. Some products are used at work, and some products are used at rest, such as an e-commerce website, most users will rest at noon. Before work in the afternoon, there will be intensive access during the period from to. For example, if the website needs to receive 1 million passengers a day, do you want to give it an average of 24 hours a day, otherwise, there will be no pressure at all. The establishment of the access model determines the correctness of the scenario design. The data can be provided by the Company's data support department to ensure accuracy.

Environment Building: the most important thing is that I often ignore this link when interviewing engineers to see if they will automatically mention it, if this step can be mentioned, it indicates that the engineer has a certain degree of knowledge about the test. The performance test environment and the function test environment are two different things. The performance test environment requires simulating the real environment. How can we simulate it here? The elements involved in the software architecture include network bandwidth, server hardware indicators, and database environment, the database content environment must be consistent with the online environment, So deploying an almost high performance environment is very important.

Tool Selection: There are many tools, and there is no best choice but the most suitable one. Therefore, it is better to choose the right one. This requirement is suitable, and there are no other indicators.

Scenario Design: There are two simple methods. One is the probe method. First, we put pressure on the tested system to find the bottleneck. Many people prefer this method. The advantage is that it is simple and easy to operate, the disadvantage is that the direction is unclear and tiring. Second, the advanced architecture requires a certain amount of Software Architecture Analysis experience. First, start with the architecture, analyze the bottlenecks that may exist in the product, then design targeted scenarios, and finally conduct unified detection. After the design, it is a good way to ask for review.

Script design: Design correct and effective scripts based on the scenario. This step is not difficult.

Test execution: the next step is used in one scenario or scenario. Record various metrics monitored during the test. The understanding of indicators represents the skill of the Performance Testing engineers. If you cannot understand those system metrics, the so-called performance testing is just a general discussion, there is no substantive significance.

Result Analysis: Based on the above test data, this analysis capability is not overnight and needs to be accumulated slowly. First, we need to familiarize ourselves with common system metrics and the significance of these metrics changes, for example, if the page/sec of the memory is frequent, consider whether the memory is not set enough to wait.

Optimization Scheme: Based on the analysis results, a reasonable optimization scheme is provided. We can start from three aspects: code-level optimization, database optimization, and architecture optimization. Code-level optimization is relatively low in cost, and the architecture optimization cost is relatively high. When it comes to this, I will go back again. In the architecture design, if some optimization suggestions can be put forward during testing, that really saves costs. Therefore, early intervention in testing is very cost-effective. Of course, the premise is that the testers are very good.

I wrote so much about it this time. After reading it, you can make a brick. You can take a picture of it, and you can take a picture of it.

This article is from humh's 51testing software testing blog: http://www.51testing.com /? 316625