Ten thousand? Twenty thousand? Fifty thousand? Or just one thousand? How many requests/connections can your system handle before it decides it needs a break? What kind of connection rates (Opened connections per second) can it handle?
These are the kinds of questions that we needed to answer in a project where we had several thousands devices connected to our backend.
In this talk I will discuss how we broke our own application by performing load tests, the problems that we encountered and how we solved them. We needed to make certain calculations and plan in a certain way in order to be ahead of the reality and make sure we are ready for the near future. I will discuss the stack we used to test these cases, namely JMeter, Docker and AWS (Amazon EC2, CloudWatch, S3, AWS-CLI). I will also discuss how to automate this in order to start your load tests via the command line or via a script.