Performance testing from scratch Implementation Guide - scene model articles

This year switched to a home appliance and commercial enterprises, performance testing from scratch. In performance tests continue to promote the process of landing, the accumulated some experience and lessons learned from scratch, he himself planned to write a series of "performance test from scratch Implementation Guide."

As already we talked to some of the things to do from scratch, such as: "performance testing from scratch Implementation Guide - Process articles" , "performance testing from scratch Implementation Guide - documents the construction of articles" , "performance testing from scratch implementation Guide - Part test plan " .

Recently busy preparing for the two-eleven full link voltage measurement related work, just received a private letter tonight peer messaging, watching feelings find many, can be considered urged the emergence of this blog it. The main contents of the private letter contains the following points:

1, performance testing, requirements analysis is the top priority - analysis is not in place can lead to scenarios not realistic, do useful work;

2, not much learning tool + monitoring costs;

3 key factors, actual performance requirements, is the impact of the final test results;

This problem points, and I feel in the actual work is very similar. I believe that many of the students performance testing, have encountered the following problems:

1, demand is not clear, sometimes even "We have a XXX interfaces, you gave me about pressure," such pseudo requirements;

2, the demand is not clear cause can not complete a detailed analysis of test points & test scenario, a large gap between the results of the final test results and the actual needs, can not locate bottlenecks and capacity planning to provide accurate online reference;

3, the positive results of the work there is no effective feedback mechanism, a problem even need this pot back pain egg thing;

。。。。。。

Of course, the actual work, will encounter many other issues, it is important: how do we analyze the reasons behind the problem, then think of ways to solve the problem!

This blog and talk about the performance testing process, and I understand how to practice the method of modeling scene. . .

 

About scene modeling, my personal practice and experience, mainly from the following three aspects:

First, the business scene

In the requirements phase before the start of the test, we must sort out the research clearly, what we test the scope and purpose?

In my Division, the main electricity supplier + is a social identification, risk control essential social services, electricity supplier's own innate logistics warehousing operations, and push-based messaging service, as well as the necessary two-eleven promotional activities, in during the two-eleven big promotion, these are the core business links.

So, we can determine an approximate range of tests, several core business includes the following links:

Know the scope of service link test, the next step is and the corresponding business & product development & students sort out their responsibility to confirm core business link (here refers to the link and the corresponding API with service attributes, as well as the importance of and priority).

Transaction business, the trading business links, will include some of the following core businesses link:

Where home might include opening screen page, login page, promotion page, personal home page; the items will include product details, product list, product collections and so on; and order, shopping cart, payment, search, and deal a strong dependence business.

According to different business segments business functions contained in the link, and the combination of system architecture type (micro poly decoupling service may do within the height based on business attributes), and prioritize the importance of different functions.

In this way, we in the requirements phase, you can get a clearer business scenario to start the next step.

 

Second, the scene link

Finished combing division of validation and test business scenarios range, then, is the demand from the test point → Split (On this point, the initial idea was to merge and pressure measurement scene is, but think about it, as an independent scene better).

In the scene link building process, the most important is to consider the following three points:

1, the task dismantling

Tasks dismantling, and as literally, sorted out according to business scenarios, from a user perspective to divide the different operating procedures, and then sort out the Task List different service link;

2, the task schedule

According to split traffic link, analyze it sort of pre-term (environment preparation, serving FBI), cross-sectoral cooperation (operational delivery, drainage channels), resource inputs (development, operation and maintenance, test), delivery of outputs ( version, API documentation, logging service, monitoring).

Then follow the time node, estimated duration and inverted, estimated man-hours to the day / person, you can have floating around for a long time, but it must be clear delivery time and plans (such as each team delivered late or insufficient resources backup plan) .

3, weight division

You see, according to the above games are played, sorted out a lot of things certain. But very often, the delivery of outputs always late, always less resource input point, the basic plan is not.

In response to this problem, as performance testing, how? In fact, the answer has been said above, the weight division and priority, for a limited time and resources into range, give priority to core test coverage and important links!

After all, the program reveal all the details, there are many, such as the flow restrictor, service degradation and even fuse (these means are lossy, but in order to guarantee service does not hang up when these are acceptable).

PS : Some of the details of things, limited to confidentiality and security rules, can not be detailed, but to seize the key, along the lines described above to practice, there is always a method for their own team.

 

Third, the pressure sensing scene

Earlier we confirmed the test range, business scenarios, business links, prioritizes and weight, made some plans and schedule tasks, but ultimately landing, or test program and the specific test scenarios.

For example, for different business scenarios, which we want to test strategy adopted, how to test, when to start, expected results and acceptance criteria, and so on.

Here, I suggest that before the final output of the test program, first draw a mind map, and the development, operation and maintenance of infrastructure and even children's shoes quickly review it, to agree on the general direction, and then enter the test program, the implementation of the pressure test.

Pressure can be measured as follows with reference to specific ideas design test scenarios:

 

As is the performance testing process, about combing the scene model building relevant content, this is mainly to share some of my personal ideas and methods, practice, it is recommended that each team based on the characteristics of targeted execution.

 

Guess you like

Origin www.cnblogs.com/imyalost/p/11415691.html