Definition of software performance testing
Definition of software performance testing: the process of simulating normal and peak load access to the system under test through automated software performance testing tools or code means to observe whether various performance indicators of the system are qualified.

Classification of software performance testing
Code-based software performance testing (the focus is on the efficiency of function or method execution)
Protocol-based software performance testing (focus on server performance)
Client performance testing (page or client response time)
Classification of server-side tests
Stress test: Under certain software, hardware and network conditions, simulate high concurrency of users (peak load) for a period of time, detect various performance indicators of the system, and pay attention to the performance of the system under the peak value (second kill, group buying, ticket grabbing)
Purpose: To monitor the operation of the system under test at the peak value, and to design a plan for the worst case (system crash)
Scenario model: gate-type scenario, a large number of threads start at the same time, and after a period of time, they end at the same time
Load test: Under certain software, hardware, and network conditions, by changing the load mode, monitor various performance indicators of the system, obtain the maximum number of users and the optimal number of users of the system under normal working conditions, and locate the bottleneck of the system
Scene Model: Arch Scene. The threads do not start at the same time, and after a period of time, the threads end one after another

Configuration test: change the hardware and software configuration (architecture configuration, parameter configuration), and observe the performance status under different configuration conditions
Benchmark test: Under certain software, hardware and network conditions, simulate a single-user operating system to test various performance indicators of the system. Make a data comparison for the in-depth performance test later.
Concurrency test: Test whether the interface works normally under the condition of high concurrency of the same module and the same application. The purpose is to mainly check whether the application or interface has defects (such as deadlock, etc.) in the case of multiple users
Capacity test: Under certain software, hardware, and network conditions, change the capacity of the database, simulate multiple users, and monitor various performance indicators. Finding the limit of data capacity
Stability test: It mainly emphasizes the stability of various indicators of the observation system under long-term and normal load conditions, and no fatal problems will occur. 7*24 hours. 8 hours, 24 hours, 48 hours. The purpose is to detect the long-term operation of the system, the stability of the system, and whether there are abnormal performances (downtime, fatal problems, etc.)

performance test using
New system: User scenarios (a large number of users, simultaneous use, and use within a certain period of time) are suitable for benchmark testing, load testing, stress testing, and capacity testing
Expansion: Analyze the performance standard of the historical system itself, and appropriately expand the capacity. Benchmark testing, load testing, stress testing, capacity testing
Tuning: As the above-mentioned system is getting slower and slower, optimize the configuration of the system to improve performance. Benchmarking, configuration testing
Fix: Solve problems such as concurrent deadlocks and memory leaks in the online system. concurrency test
Lightning kill, group buying: benchmark test, load test, stress test
The time to intervene in the performance test is generally after the function and interface tests have been completed before performing the performance test.
Metrics for performance testing
Response time: the period from when the user sends a request to when the server finishes processing the request and returns it to the client
Throughput: Measures the business processing capability of the system. TPS: Transactions per second. QPS: the number of requests per second
Resource utilization: cpu, memory, network, disk read and write io. The utilization rate of general resources is not higher than 70%-80%. If something is higher than this value, it may be a performance bottleneck
Error rate: The probability of a failed request when the system is under load. Error rate = (number of failed requests/total number of requests)*100. Different from errors in functional testing, in performance testing, the so-called errors generally refer to errors caused by system timeouts, rather than functional errors. Different systems have different error tolerance rates. For ordinary business systems, the error rate should not exceed one in ten thousand, and for some large systems, one in billion.