Your 44% link is after crippling the benchmark by de-parallelizing it on top of using a poorly written Erlang implementation. I suggest you read the actual blo

Your 44% link is after crippling the benchmark by de-parallelizing it on top of using a poorly written Erlang implementation. I suggest you read the actual blog post that your Google+ post links to, including its comment section and the links posted therein (which is how I arrived at my links): http://www.krazykoding.com/2011/07/scala-actor-v-erlang-genserver.html http://www.krazykoding.com/2011/07/scala-actor-v-erlang-gens... > Is 100 microseconds not short enough for your application? GC time is not a useful benchmark by itself. The important thing with the JVM is that you can't predictably reason about when the global GC will happen (global GC pause). Your JVM actors aren't shielded from these global GC events so building a soft real-time system is less practical under the JVM than the Erlang VM IMHO.