1. Basic concepts of performance testing
Performance testing: Refers to the use of automated testing tools or code to simulate a variety of normal, peak and abnormal user load conditions under high concurrency through multithreading, and observe whether the system's various performance indicators and monitoring conditions meet the standards.
Load test: Check the performance of the system under a given load, resource utilization, and whether it reaches the expected performance indicators, etc. The bottleneck of the system is usually not emphasized. Usually need to test the performance under the existing load and the expected load, and conduct a comparative analysis.
Stress test: Continuously exert "stress" on the specific software and hardware environment, test the performance of the system under stress, and obtain the maximum that the system can provide by determining the bottleneck or unacceptable performance point (system crash point) of a system Service level testing.
Capacity test: Under the established environment (software environment, hardware equipment, network conditions, basic data, etc.) and meet certain conditions (such as transaction response time), obtain the maximum processing capacity of the application system.
Stability test: By loading the system with certain business pressure, let the application run for a period of time to test whether the system can run stably under this condition. Generally, let the system be under the peak pressure for 8-10 hours of testing.
Large data volume test: It is usually a comprehensive test of large data volume for certain system storage, transmission, statistics, processing and other services, such as big data platform scheduling processing, mainly focusing on whether the processing time window meets the requirements.
Failure recovery test: For the system under test with redundant backup and load balancing, it is checked whether the user can continue to use the system if the system fails locally; and if this happens, the user will be affected to what extent (such as transaction response time) Is it within the acceptable range).
Configuration comparison test: By adjusting the system software/hardware configuration, understand the impact of different environments on system performance, and find the optimal allocation principle of system resources, which is generally used for performance tuning and system planning.
Second, the performance test process
Three, performance test indicators
Number of virtual users VU: The full English name is Virtual user, also called the number of concurrent users. Multithreading is used to simulate the pressure caused by different users on the continuous access to the system.
TPS: The full English name is Transactions per second. It is an important indicator to measure the processing capacity of the system. A page, an interface, and a business flow can all be a thing.
Average response time ART: The full English name is Average response time. Response time is an important reference indicator to measure the processing speed of the system. It is related to the user's intuitive experience. Generally, the statistics are the time consumption of pages, interfaces, transactions or a complete process.
Success rate SR: The full English name is Success Rate. The success rate is an important reference indicator for measuring system stability, and it is generally required to be above 99%.
Resource utilization rate: CPU, memory, disk IO, network IO, etc. The utilization rate generally does not exceed 80%.
Preconditions: based on what business scenario, how much data, and how many concurrent users.
Response time analysis:
Response time = network delay + system processing time; ResponseTime=(N1+N2+N3+N4)+(A1+A2+A3);
Response time "2-5-8 principle":
