JMeter test equivalent in NBomber

[botondberes]

I'm trying to make a case for NBomber over JMeter but I have trouble getting my scenario to behave as the test in JMeter.

Can someone please help me figure out what this JMeter configuration translates to in NBomber?:

My assumption was that this would translate into these calls:

            .WithoutWarmUp()
            .WithLoadSimulations(
                Simulation.RampingInject(rate: 100,
                                        interval: TimeSpan.FromSeconds(1),
                                        during: TimeSpan.FromSeconds(10))

But that doesn't seem to be it because after running this, JMeter shows over 54k requests at over 5k RPS as you can see below, whereas NBomber shows 450 at 45 RPS:

Any help is appreciated!

[AntyaDev]

Hi @botondberes ,

Can you please try

Simulation.KeepConstant(copies: 100, during: TimeSpan.FromSeconds(10))

and also

Simulation.Inject(rate: 100, interval: TimeSpan.FromSeconds(1), during: TimeSpan.FromSeconds(10))

[AntyaDev]

@botondberes
By looking at your Jemeter I think you shouldn't use RampingInject since it will inject users slowly (over 10 sec) in your case.
https://nbomber.com/docs/nbomber/load-simulation#ramping-inject

[botondberes]

@AntyaDev Thank you very much for your quick response. KeepConstant seems to be the closest. I assume it kicks off with 100 copies immediatelly instead of ramping them up over a second compared to JMeter but it's close enough. To be fair, ramping up to 100 over a second doesn't really seem realistic anyway. Running JMeter with 0 seconds ramp-up is essentially the same as KeepConstant and shows that NBomber is about 9.5k requests faster than JMeter.

Could you answer one more question for me?
If I have two scenarios registered in NBomberRunner, is there a flag to make those run sequentially instead of parallel or do I have to create two runners and feed each it's own scenario instead?

[AntyaDev]

Unfortunately, there is no such flag to run them sequentially.
Do you want to run them sequentially and get a single HTML report containing both of them?

[botondberes]

Yes, ideally it would be one report containing them both. The idea here is that if both scenarios target the same host, it might not be beneficial to have them run in parallel. For example if the tests are not for load or spike testing but functionality validation and testing how long it takes for requests to be processed under normal circumstances (get a baseline).

[AntyaDev]

Hi @botondberes
Sure, it sounds reasonable.
I would suggest submitting a dedicated task for this feature.
Also, if your company has a license for NBomber then such tasks usually get higher priority.

[botondberes]

I think the sequential feature request already exists here: #494

Thanks for your quick responses. We're in the process of deciding which way to go. I'll keep you posted!

[AntyaDev]

Hi @botondberes
thank, I have added a task for this: #643