First look at the pressure test report form
Mainly find the problem from the following dimensions:
(1) Check whether the error rate is within the allowable range.
When there are many stress test threads, there will be an error rate, even a simple return 0; will have a response timeout. We can judge whether the corresponding error rate is allowable under this pressure environment. At the same time, we can optimize the code and optimize SQL to improve response speed.
(2) Whether the TPS throughput meets the needs of business scenarios.
During the stress testing process, keep an eye on TPS changes, check the approximate maximum TPS throughput, whether it meets business needs, and record the highest TPS concurrency number and the optimal concurrency number. If the number of TPS processing per second is greater than the number of concurrent threads, you can continue to increase the number of threads. If the number of TPS is less than the number of concurrent threads, you must reduce the number of concurrent threads.
(3) Keep the memory and CPU usage of the server under pressure for comparison.
Find problems in time and optimize code quality (for example, whether excessive Full GC is generated, timely adjustment of memory allocation, whether there is slow SQL, whether there is an endless loop generated by certain conditions, whether the creation of large objects and containers is reasonable).