[erlang-questions] Messaging mechanisms between Erlang runtimes?

[Allen McPherson]

I'm running network performance tests on clusters with
distributed Erlang. I'm trying to analyze what I'm seeing without
asking a bunch of questions here (or digging through the
source).  When I get most of it wrung out I'll attempt to
post code and results, but for now I am seeing some strange
stuff...like similar messaging bandwidth regardless of whether
the endpoints are on the same node or different nodes. There
are all kinds of variables I need to work my way through,
but one quick question I have is this:

Does Erlang implement any optimizations for messaging
between runtimes on the same node? My code is pretty
simple: slave:start_link to local (localhost) or
remote nodes, spawn a bunch of tasks on these
nodes, and time large message traffic between them in
various patterns (ping-pong, all-to-all, etc.). Note
that on the local node this is not messaging between
processes (Erlang "threads") but between different
instances of the Erlang runtime. Obviously, messaging
within a single runtime blazes (especially with
term_to_binary).

Thanks
--
Al