Ali cloud performance testing tool reference documentation https://help.aliyun.com/document_detail/29338.html?spm=a2c4g.11186623.6.615.440c14761SSHDt
concept
HPS
hit per second, hits per second
QPS
query per second, the number of queries per second
TPS
Transaction Transaction Per Second, the system can handle per second
RPS
request per second throughput, the number of requests per second
TPS is generally used to measure the business, QPS interface to measure the number of queries, HPS measure clicks request to the server
Standard (the bigger the better)
- The financial industry: 1000TPS ~ 50000TPS, not including Internet-based activities
- Insurance industry: 100TPS ~ 100000TPS, not including Internet-based activities
- Manufacturing industry: 10TPS ~ 5000TPS
- Internet e-commerce: 10000TPS ~ 1000000TPS
- Medium Internet website: 1000TPS ~ 50000TPS
- Small Internet site: 500TPS ~ 10000TPS
Response time
Client sends a request starts, the client receives the server returns the end (request transmission time, network transmission time, the server processing time composition)
Performance test results, are divided into minimum response time, maximum response time, average response time
Standard response time
- Internet companies: 500 ms or less, for example, Taobao business is about 10 milliseconds.
- Financial companies: less than one second better, part of a complex business than 3 seconds.
- Insurance companies: 3 seconds or less is better.
- Manufacturing: 5 seconds or less is better.
Error rate
Error rate refers to the system under load, the probability of failure of the transaction. Error rate = (number of failed transactions / total number of transactions) * 100%. Better system stability, the error rate should be caused by a timeout, timeout is the rate.
standard
Generally does not exceed six thousandths Serve power of not less than 99.4%
CPU and system load
CPU resource usage test scenarios servers running period, factors to determine whether the stable operation and processing capabilities
Measurement system was working and how much: the system load
standard
PU indicator mainly refers CPU utilization, including user mode (User), the system state (SYS), wait states (the wait), an idle state (idle).
CPU utilization in the industry to be less than the warning value range, i.e. less than or equal to 75%; CPU sys% less than or equal to 30%, CPU wait% less than or equal to 5%. Single-core CPU also need to follow the requirements of these indicators. CPU Load is less than the number of CPU cores.
Network Throughput
The number of data in case of failure without network through a network per unit time. Unit is Byte / s
When the network throughput targets close to a network device or link maximum transmission capability, you need to consider upgrading network equipment
standard
There are number of megabytes per second flow out, can not exceed 70% of the maximum transmission capacity of device or link under normal circumstances
Disk Throughput
In unit time without disk failure case the amount of data disk
standard
Disk busy rate is a direct reflection of an important basis whether there is a disk bottleneck, under normal circumstances, the disk busy rate is lower than 70%.
Complicated by
Narrow: multiple users at the same time the same operation. In order to test the database / program processing concurrent operations for load testing, stress testing
Generalized: a plurality of different user operation, for the system, or operating many users colleagues, for mixing the scene, stability testing
Scenes
In order to simulate real user's operation, the scene includes a script, pressure pattern, number of users, log level, the pace of time
Number of concurrent users
Simulate the real steps of virtual users
Request Status
200: success, 204: success but did not return anything, 3 **: the need for further action in order to complete the client request ** 4: The request may be wrong, the server can not handle 5 **: internal error occurred while processing the request server