String manipulation is notoriously slow in the BEAM, mostly implementation reasons. Message passing is somehow slow as well, slower than go(Lang) of you want a

String manipulation is notoriously slow in the BEAM, mostly implementation reasons. Message passing is somehow slow as well, slower than go(Lang) of you want a comparison. But on those systems, the one that you design on the BEAM, velocity is usually not a problem. What you usually try to do is to reach a design that does not have a single bottle neck or failure point so that pretty much whatever happen you can just add machines. This turn out to be a great way to design multi{processes, cores} and multi nodes system. What you usually get on a BEAM under load is an 100% of CPU usage, even on multi core, but to be honest those core are not used as efficiently as you could. It is a matter of tradeoffs, I can quickly write multi process software that can easily scale, but I will leave on the table some raw performance number. Again it turns out that just raw numbers are not as important as they are simple to measure.