Outline
jmeter provides many listeners performance data, we analyze the performance bottlenecks by listeners can
Taking a step pressurization test results 500 threads to describe the chart.
Common Listener
1：Transactions per Second
Monitor dynamic TPS, to analyze throughput. Wherein the abscissa is the running time and the ordinate is the value TPS. Red by the TPS, green indicates failure.
TPS maximum of around 140, from about 1 minutes 26 seconds, began things does not pass
2：Hits per Second
CTR dynamic monitor unit time, that is, the number of requests triggered. Wherein the running time is the abscissa, the ordinate is the value of HPS.
Click-through rate fluctuations, and can not continue to rise. Description Performance is very unstable
3：Response Times Over Time
Listener response time during the entire run of things. Wherein the abscissa is the running time and the ordinate is the response time (in milliseconds)
Response time is about 4950ms began to stabilize, the follow-up has undergone a major fluctuation
4：Response Times vs Threads
Response time during thread activity monitoring. Wherein the abscissa is the number of active threads (i.e. concurrent), the ordinate is the response time (in milliseconds)
5： Active Threads Over Time
The number of active threads within the monitor unit of time. Wherein the abscissa is the unit of time (milliseconds), the ordinate is the number of active (i.e. concurrent) threads
6：Response Times Percentiles
The percentage of monitor response time distribution. Wherein the abscissa is the percentage of requests, and the ordinate is the response time. This figure represents 99.7% of the request response time of less than 5s.
7：Response Times Distribution
Response time distribution histogram. Wherein the abscissa is the histogram of the ordinate is the response time. The figure shows approximately 111 request response time 5076ms.
8：Composite Graph
Combined listener. Wherein the abscissa is the running time and the ordinate is the values of the aggregated performance data (some of which data is divided by 10).
to sum up
Different listener can listen to different performance data, but want to chart out visual analysis of performance bottlenecks, we need to modular listener. For example, the throughput obtained by the distribution of response time and throughput inflection.
通过以上图表能看出来，在持续加压的事物场景中，99.7%的请求响应时间都控制在了5s以内。