What's the relationship between QPS/TPS, response time and number of concurrent users

Question

Some Concepts:

• TPS means Transactions per second
• Response time is the total amount of time it takes to respond to a request for service

Is this formula true?

TPS = number of concurrent users / response time

Answer 1

It is true if transactions happen sequentially and in only one thread (on one TCP connection) per user. In reality, however, when talking about web browsers, they will use multiple concurrent connections when talking to a host. 6 concurrent connections is quite common, so the host will then get TPS = 6 x concurrent users / response time.

Also, the browser will sometimes be blocked and not fetch things. Sometimes because it is executing code, sometimes because it cannot perform some operations simultaneously with other operations. See http://www.browserscope.org for more info.

Also, of course, clients (whether they are humans using a browser or e.g. a mobile phone app talking to its backend via a REST API) don't usually make requests back to back, continuously, at the highest possible rate. That is probably not a very realistic test case. Usually, clients will make a bunch of requests and then fall silent for a while, until the user does something new in the application that requires more data from the backend.