1. Before you begin, first introduced under pressure measured some basic plug-ins: common thread group into three categories: user thread, step thread, ultimate thread:
user thread: the most versatile of the most original thread implementation; into circulation implement threads, the thread can be achieved delay delay;
step thread: to achieve some of the more complex scenes, such as common climbing class type, and continuous online scene
This Group will start 10 threads: this test will play a total of 10 threads. First, wait for 0 seconds: Wait 0s start thread, which is not directly available from the thread. Then start 0 threads; start continued to increase from 0 threads. Next, add 2 threads every 3 seconds : each additional thread runs after two 3s, renewed the remaining two threads run 3s, and so on. Using ramp-up 6seconds: in front of two threads each time to spend from 6s, namely from up above in connection with two threads within 6s, running 3s, then amassing renewed two threads within 6s, then run 3s, and so on. Then hold load for 30 seconds:. After all the threads together, began to stop running after 30s. Finally, stop 2 threads every 1 seconds : finally stop the thread, two threads to stop once, then 1s, etc.
ultimate thread:
Definition explain: Start the Count Threads: start of the current line total number of threads Initial Delay / sec: delay start thread of the current line, unit: second Startup Time / sec: The time required to start the current line of all lines Cheng Dafeng value, unit: seconds Hold load For / sec: current line thread stable after the peak load time, unit: s Shutdown time: time required to stop all threads in the current line, unit: seconds
2. About synchronization timer syn timer:
It has two parameters: 1. The number of simulated users group I have here called him the set of release threshold means that when you want to implement the user reaches a certain amount with simultaneous requests purpose, he will decide what time to send according to your timeout time setting has been assembled the request requests
You need to pay attention to are the following: 1. The number of simulated users value can not be greater than the number of threads in the thread group user thread of the fill value, and secondly for the syn timer timeout 0 indicates the timer will wait until the number of simulated users reached before setting all the requests will be issued once, when non-zero, such as over set time has not yet reached the set number of requirements, not waiting for a request has not reached the back, directly transmits all requests arrive;
2. If your number is the number of simulated users group gathered defaults to 0, he will be in accordance with the number of use thread thread waits for the timeout studied a little trick is to time> number of simulated users group * 1000 / user thread number / loop count can be avoided because the set timeout = 0, the appearance has been waiting for to simulate user group the number of cases stuck phenomenon
3. About Plug-ins: tps trt, activate thread, monitoring and graphical summary plug-Download: Extras.jar
https://github.com/chen1932390299/JavaProject/raw/JmeterJavaBranch/lib_jtlChange/ext/JMeterPlugins-Extras.jar
4. Data analysis: For many people stress tests do not directly consider the case of sustained pressure test to a relatively short time period of data to measure the overall service performance data is very scientific:
First, what is the number of virtual users, what is the amount of concurrency, and even some specific differences pm expression and did not find out the number of concurrent users themselves or when the user needs, jmeter user simulation is achieved by a thread, a thread represents a virtual user; many novice one up is the number of threads equal to the number of concurrent heap up, that is dry and some even took a direct windows
Simulate 5000, even higher data concurrency is a lot of experienced technical staff criticized it can achieve the desired effect, then there is no thought why not do it, first of all can cotinine 1