Lao Li shares knowledge: TPS and throughput rates for performance testing

When the pressure of the system is increased (or the number of concurrent users is added), the throughput rate and the change curve of TPS are broadly consistent, the system is basically stable.

If the pressure increases, the throughput curve added to a certain extent after the change is slow, even flat, while TPs tends to flat, view system resources utilization, if the resource utilization rate is low, indicating that the server hardware resources are not in doubt, view network traffic, estimated network bandwidth there is doubt.
Similarly, if the CTR/tps curve changes slowly or flat, click-through rate (number of requests per second) if the pressure is added, tend to flat, it is likely that the server response time to add, observe the server resource utilization, to determine that can not be a server query.
TPS is the abbreviation for Transactionspersecond, which is the number of transactions/sec. It is the unit of measurement for software test results. A transaction is a process by which a client sends a request to the server and the server responds. The client begins to time the request, receives the server response, and ends the timer to calculate the elapsed and completed transactions, and at the end uses the information to estimate the score. The client uses the weighted covariance averaging method to calculate the client's score, and the test software uses this information from the client to calculate the overall TPS score on the server side using the weighted covariance averaging essentials.
Generally, evaluation system performance is measured in the number of skill trades completed per second. The overall processing capacity of the system depends on the TPS value of the lowest processing capacity module. According to experience, the ability to use the system usually requires about 10-100 of the processing capacity. There are significant differences in TPS for different systems, and it is often necessary to perform accurate estimates through performance tests.

Original link: http://www.cnblogs.com/laoli0201

Lao Li shares knowledge: TPS and throughput rates for performance testing