Image results:
- By observing the average sampling response time, the user can visually see the changes in system performance as the concurrency pressure increases and the performance test time increases. Normally, the average sampled response duration curve should be smooth and roughly parallel to the bottom edge of the image.
- Exception 1: The average response time jumps in the initial stage, and then gradually smooth down. It is indicated that the system has a slow effect in the initial stage, and the cause of this phenomenon is that the system has a performance defect in the initial stage, and Xu is further optimized; the second is that the system has a caching mechanism, and the performance test data is not changed during the test, so that the same data will respond slowly in the initial phase. And the subsequent response time is basically consistent, this is the performance test data preparation problems, not performance defects, need to adjust and then continue to test; Thirdly, the inherent phenomena caused by the system architecture design.
- Exception 2: The average response duration continues to grow, and the graphics become steeper. This is certain to have a performance flaw (such as memory leak i), which can be used to locate the problem by monitoring the system logs and monitoring the server performance metrics.
- Exception 2: Average response duration during performance testing, a sudden jump occurs and then returns to normal. The cause may be due to system performance flaws, and may be caused by instability in the test environment. Troubleshoot system environment, network, etc.
- The uniform distribution of sampled data can be judged by observing the standard deviation of sampling response time. When the sampled data is unevenly distributed, there may be a performance flaw. The standard deviation reflects the degree of dispersion of a data set. The higher the standard deviation, the more discrete the test data, and the lower the standard deviation, the more stable the data representing the test. The ideal sampling response time-length standard curve should also be smooth.
- The actual number of samples processed per minute by the throughput standard server. Find the maximum throughput supported by the system by increasing the number of concurrent threads or by reducing the latency in the script.
JMeter Study Notes---performance analysis