Hacker News new | past | comments | ask | show | jobs | submit login

What in particular about gen_server:call is not working for you?

All I can really do in reply to this is ask you to reread the post I wrote more carefully. If you have a more specific clarification, let me know.

I read it, I'm asking pragmatically; although in theory using worker pools can have problems at the edges, in practice erlang's message processing speed is quite high, and, e.g., worker_pool in production at many thousands of messages per second is working great for me. The latency of a single message pass is submillisecond. You're having problems, despite that?

I covered that. If you have high variance in how long message processing takes, which is very easy to do accidentally, then if you something simple like "random" you can easily block up the message latency. (Of course, "don't do that". Easier said than done.)

If you're at a scale at which you're seriously using Erlang, "fast" isn't really a concept that applies anymore. You really end up with "enough performance" or "not enough performance", with an at-times surprisingly sharp transition from one to the other, and while the base performance of the underlying language is certainly a major factor in whether you have "enough", it isn't the only factor, or even necessarily the most important. (Which is good for Erlang, because it's actually quite slow at most conventional things.) It doesn't matter if I've got "sub-millisecond message processing" if I'm trying to run more messages through a box than it can handle, or if I've got an external response handler that sometimes clogs up.

There's no way that Erlang can possibly be "fast" enough that somebody won't throw a task at it that's still a performance challenge.

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact