Thanks 🙂
There are thread pool names in the logs, and you can find information about it in Spring documentation.

I’m just configuring a dedicated thread pool, to use a different pool then the main worker thread pool

when u write

return Flux.fromStream(this::prepareStreamPart1).log()
.mergeWith(
client.get().uri(“/slow/” + param)
.retrieve()
.bodyToFlux(Person.class)
.log()
.publishOn(Schedulers.fromExecutor(taskExecutor))
);

as you have used publishOn in the end, how will it benefit at all? The webclient will still use nio thread group to call and receive responses over network. So, what purpose does publishOn serve here?

And how can performance increase with this change as you have shown in your performance results?

Yes, of course. They are workers.

Yes, ‘epoll*’ is for that fragment `Flux.fromStream(this::prepareStream).doOnNext(person -> LOGGER.info(“Server produces: {}”, person))`, while ‘nio*’ is for the `WebClient`:
“`
client.get().uri(“/slow/” + param)
.retrieve()
.bodyToFlux(Person.class)
.log()
.publishOn(Schedulers.fromExecutor(taskExecutor))
“`
So, the client and server use different thread pools.

Well, there is a lower number of used threads, and probably a little smaller memory footprint (I didn’t test it). In this article I focused on showing threading model, not on performance testing. However, if you will increase the number of concurrent threads in the test you should have better result in comparison to synchronous approach.