Users per second and response times

Gatling (Open-Source)
1

Hi,

Am I correct in saying that a user waits to be be notified of a response
through a future?

Basically :slight_smile:

So, if I have 50 users per second in the simulation, but the total
response times for that user are > 1 second, will I be unable then to reach
50 users per second in the test run?

That's an open model, user arrival rate is not related to the response
time. New users keep on arriving at a 50/s rate, period.

2

Your question was about user arrival rate, not rps. :slight_smile:

If scenarios contains more than one request (ie each user performs more than one request), of course global rps will drop if response time increases:

  • 1s response time + 5s think time + 1s response time = 2/7 rps = 0.29
  • 5s response time + 5s think time + 5s response time = 2/15 rps = 0.13
3

Each individual virtual user behaves like a physical person: you have to wait for a response (like a web page) before being able to perform the next action (like clicking on a link in this page).
But virtual users are NOT coordinated altogether.

4

Requests per second or responses per second?
How high do you load? What does Gatling GC activity look like?

5

I ran jstat and didn’t see any unusual gc activity when the requests dropped off (at around 525 RPS). I duplicated the request in the scenario and ran my desired load at 800 RPS without an issue. Not entirely sure why this would be the case.

6

I duplicated the request in the scenario and ran my desired load at 800 RPS without an issue.

Sorry, I don’t get what you mean. You mean that you ran the simulation a second time?

7

Would need a reproducer.

8

If scenarios contains more than one request (ie each user performs more than one request), of course global rps will drop if response time increases:
How can I keep the global RPS consistent?