Hi all
I am looking for opinions and discussion based on experience in running relatively large user counts… very roughly between min 500k and 10m max users?
I’m not sure if this is even appropriate for this tool? A whole lot of unanswered questions, that are unable to be answered without hands on experience.
I have submitted an official ticket but no harm done here!
Looking forward to your responses.
Brad.
What is the purpose of simulation of 10 mln vusers? What is real rps you want to produce? I suppose it depends on your load profile and the character of load - do you use correlations, data checks, execution of complicated custom method during your test and so on. Anyway you’ll need multiple load generators for such huge load. But it looks like you need to implement hit-based load approach from my point of view.
суббота, 4 декабря 2021 г. в 10:44:56 UTC+3, bradle...@gmail.com:
10M concurrent users is huge. Are you sure about this target?
You’ll need a very large number of injectors to be able to generate such load, because of some hardware, operating system and protocol hard limits:
Number of connections: you can only open ~65k TCP connections per injector to a given remote IP address (simplifying).
Ephemeral ports recycling speed
Bandwidth
CPU for processing IO, decompressing, handling encryption, etc…
Depending on your use case, you might need 150 injectors or more to be able to generate that kind of load.
Regards,