Hello Bruno, thank you for the response.

No, I do not have any experience in Kotlin, I might take a look at it :)

For the rest, I beg to differ.

When talking about high concurrency (and not parallelism), the most interesting metric is the total throughput.

In addition, the number of virtual users that the JVM can handle with 1 CPU, if my tests are correct, is like 1500, which is a lot. More than 1500 will result in not-blocking implementation to be faster, but only having 1 CPU available and with an average response time of 45 seconds....

and with that high time, you should consider increasing the CPU allocated to that service, or increasing replicas... and that's where JDBC becomes better again.

Take in mind that here we are talking only about REST web calls, and very specific implementation.

Furthermore, java does not create a thread for each request but uses a thread pool. In the latest versions of java memory used for thread creation has largely decreased.