LoadRunner did not tell you-2-descriptive statistics and performance result analysis (continued)

Copyright Notice: This article can be reproduced, but it shall not be used for any commercial purposes or other purposes for profit without my permission. I reserve all rights to this article. If you need to reprint the content, Please retain this copyright statement and ensure the integrity of this article. You can also understand the hard work and respect the work of the author.

Author: Chen Lei(Jackei)

Email:Jackeichan@gmail.com

Blog:Http://jackei.cnblogs.com

 

The idea of statistical analysis is equally important as the presentation of analysis results. With good analysis ideas, we do not know how to better display the analysis results and data to prove our analysis, it's like a person who is full of economic resources, but doesn't know how to show his talents.

Pai_^

 

this time, I will use two charts to describe other applications of" descriptive statistics "in performance test result analysis.

In this figure, we continue to use the previousArticle-- The method in descriptive statistics and result analysis analyzes the distribution of response time. The data used in the figure above is

Google.comThe homepage is tested and used separately in the test.10/25/50/75/100Number of concurrent users at different levels. Through this chart, we can use horizontal comparison and vertical comparison to better understand the responsiveness of tested applications under different levels of load.

 

This chart uses the same data as the first one, but we use another perspective to present the data. 2000/5000/10000/50000 the unit is millisecond, indicating that during the test, the response time is 0-2000 Percentage of successful transactions in milliseconds, the response time is 2001-5000 Percentage of successful transactions in milliseconds, the response time is 5001-10000 transactions in milliseconds Percentage of successful transactions, the response time is 10001-50000 Percentage of successful transactions in milliseconds.

the determination of these time ranges must refer to the industry's" 2-5-10 principle"-Of course, you can develop other standards for your own tests, as long as you get recognized by the enterprise. The so-called " 2-5-10 principle", simply put, that is, when the user can 2 when a response is received within seconds, the system will quickly respond; when you 2-5 when a response is received in seconds, the system's response speed is still acceptable; when you 5-10 response within seconds, the system will feel slow, but acceptable; and when the user exceeds 10 seconds later, the system will feel bad, you can also choose to leave this Web site, or initiate a second request.

As shown in the preceding figure, when the number of concurrent users is10, Exceeds95% Of users can5Response within seconds; when the number of concurrent users reaches25, There are already80% Of the transaction response time is at a critical risk, and a considerable number of transactions have a response time that exceeds the user tolerance. As the number of concurrent users increases, more and more transactions exceed user tolerance. When the number of concurrent users reaches75The system has almost no response to any user.

This chart can also be used to compare and analyze the transaction Success and Failure ratios under different loads.

 

Note: The data in the above two charts is mainlyExcelProvided inFrequency,Average,Max,MinAndPercentileSeveral statistical functions are obtained. For detailed usage instructions, seeExcelHelp manual.

Click here to learn about the creation progress, view the document directory, or browse completed articles.