I recently used gatling.io for load testing my api server. During testing there were options to set for example, number of concurrent users per second.
What I do not understand is how these tools are able to generate that load. How are they able to create lets say 200 concurrent connection to my server and hit it?
I am looking for an answer describing how it is implemented, whether they use thread, socket programming etc. Thanks.
Gatling uses the Akka framework internally where every virtual user is represented as an Akka actor.
Akka is a framework for asynchronous execution. This means we do not waste and suspend the thread waiting for the response but use it to complete other tasks - in this case of other virtual users. The time of execution is managed by the scheduler and execution itself by the dispatcher
Considering in some environments a single JVM thread stack can get as large as 2 Mb or even bigger creating a thread for each of 500 users would cost about 1 Gb of memory just for the threads. See here for more information about why creating lots of threads is a bad idea.
In conclusion, it is essential to use asynchronous programming to achieve such a high throughput of concurrent virtual users.
