Shulou(Shulou.com)06/02 Report--

Flow Control Test method for Service side performance Assurance

On the last Sunday at the end of July, the second phase of our offline performance improvement class was officially opened, with many students with zero foundation, and several people with test management experience and many years of development or testing experience. but each is very eager to learn, not because of learning, but with a common ideal to walk in a classroom to exchange and discuss professional knowledge. They are modest and self-motivated, because they know they are doing non-cumulative work, but expect the boss to give you a raise? Even if they have done it for a long time, they may not be able to get much, and everything depends on their strength, so they cherish every class, take careful notes and ask divergent questions, regardless of functional testing, test management, industry knowledge, architecture design, software and hardware knowledge, and so on. it is indeed a test of teachers' knowledge and practical experience.

Students at different times, in different stages of learning promotion, students ask shallow questions of different depth, depth and breadth, all over the world, which is really interesting. I will also improve and enrich my knowledge system in the teaching exchange, and there will also be some problems. I will not consider so perfect in my usual work, but also because of gaps in knowledge, let me make up after class, let me gradually improve my management system and professional knowledge in various fields. And applied to the actual project work, for example, in the second class, students asked about the differences and usage of char and varchar in the performance of database table design, the performance impact of different types of disks, the test of APP weak network, the causes of abnormal transmission between southern and northern wireless networks, database attempt scenarios, how to estimate the number of concurrency and online, and so on. Different technical questions have different argumentation methods and can be written into a technical article.

This chapter introduces the system flow control or current restriction, and it is also a question asked by a student in class, which I think is worth discussing, because in 2012, I tested a large state-owned project. The system has complex processes, many rules to verify, a large amount of data, a large number of concurrent users and a high degree of concentration and concurrency. Customer requirements and project operation and maintenance technical problems, really have deep experience. In addition, in the era of Internet trading, all kinds of APP products are everywhere, and various marketing methods emerge one after another in order to seize the market and user traffic. at this time, we can not use normal means to evaluate the number of users of the system. In order to improve the reliability and user experience of the system, software development will improve the performance of the architecture from different architectural design angles, non-functional design and so on. This is also one of the reasons why micro-services are becoming more and more popular, through micro-service architecture to improve the high availability and high reliability of the system under high concurrency.

Product needs flow control background

For example, the emergence of various e-commerce APP, it is indeed difficult to predict the number of online users, registered users, concurrent users, and even tourists querying product information users will exert pressure on the service in the process of performance testing. this is really difficult to estimate, so how do we do it?

Now all kinds of e-commerce, Meituan, ele.me, Taobao, JD and other high concurrency performance pressure scenarios such as second kill, big promotion, double 11,618, etc., the flow control of the server has become a rigid architecture design index requirement, in order to ensure that the system can run smoothly under high pressure to play the role of high availability and high reliability.

In fact, many external interface services are the same. In order to prevent the external request data from being higher than their own estimate, the external service interface will set the interface current limit effect. Especially for the micro-service system, the interface request may come from many external system calls. For this kind of micro-service with massive interface requests, interface current restriction is very important.

Which indicators should be used for flow control?

The performance metrics involved in our performance testing process, such as throughput, TPS, HPS, and the number of concurrent users, can all be used as traffic control metrics. Of course, some business metrics can also be used, such as time limit for entering CAPTCHA, input times of SMS CAPTCHA, institutional user limit, JSF request limit, and so on.

In general, it is convenient to control the number of requests, that is, HPS, because the value is searchable and controllable, and it is easy to pass the performance stress test.

Scene case introduction

The following scenario is that I assisted a relatively large domestic city commercial bank to conduct a flow control verification test of the credit system in 2012.

Project background:

The main purpose of the traffic monitoring test of a certain system is to verify whether the stability test when opening the monitoring tool has an impact on the use of resources such as system memory when grabbing response timeout JSF and UCC; at the same time, when turning on the traffic access control switch, it accesses specific JSF and UCC according to a certain number of concurrent users of the system, and verifies whether functional control is carried out or not.

Purpose of the test:

The flow control user access restriction test mainly focuses on whether the front end prompts the user number access restriction information prompt when increasing the number of users under the pressure of the system to access a business transaction point under the pressure of a certain number of concurrent users.

Test strategy:

The main test strategy is to verify whether the pop-up access volume exceeds the maximum access prompt information controlled by the traffic switch when the number of UCC requests exceeds a certain number.

Test results:

After setting the business transaction control scenario and performance test scenario for traffic start, set instant concurrency to operate a business UCC service request. When the request is exceeded, the following information should be prompted to ensure that the system is successfully controlled by flow.

And in the stress testing process to check the memory recovery, server resource utilization, session session recovery to ensure that the number of sessions is correct and the recovery rate is accurate.

In the loadrunner stress test scenario, you can also see the following error message, indicating the accuracy of the flow control function under concurrency.

The above scenario is for traffic control for business requests, which can be regarded as business flow control, but in the general micro-service era, traffic monitoring may be more complex when the technology interface and cluster deployment mode are used. For example, non-functional tests are carried out to verify the frequency limit and scheduling frequency of each interface service.

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.