2024-07-12
한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina
Can the system run stably under a large number of users (more)
For hardware server selection
For selection of software technology
User perspective: response time
Resource usage: number of concurrent users, TPS, resource usage (cpu, memory, JVM)
Benchmark test: Single-user test, conduct separate tests on important services in the business model to obtain various performance indicators when a single user is running
Load testing: testing the changes in system performance by gradually increasing the load, and ultimately determining the maximum load the system can withstand while meeting system performance indicators.
Stability test: long-term testing when the server is stable
Response time, concurrency, throughput, error rate, resource utilization
QPS: The number of requests processed by the server per second (that is, the server interface)
TPS: The number of client requests processed by the system per unit time (one operation on the interface, one request or multiple requests)
The probability of failed services when the system is under load. Error rate = (number of failed services/total number of services)*100%
Generally no more than 1 in 10,000 (as specified by the company)
The error rate of a system with better stability should be caused by timeout.
It is recommended that the CPU should not exceed 80% (plus or minus 5)
Memory is no more than 80%
Disk is not higher than 90%
LoadRunner: An industrial-grade standard performance load test that can simulate tens of thousands of users to implement tests
Jmeter is free and can provide basically the same functions as LoadRunner, meeting the needs of most companies.
Start jmeter
Add a thread group under the test plan
Add http request and sampler under thread group
Fill in the relevant request data of the http request
Add View Results Tree and Listener under Thread Group
Click the Start button to run and view the results