Ignoring payload size, what is relative difference in latency between an in-proc call (say in C++ or Java), a socket call to a process on the same machine, and a socket call to a process on another machine? This can be expressed as a minimum latency in ns/ms or in terms of relative orders of magnitude.
I'm looking for something similar to this:
http://duartes.org/gustavo/blog/post/what-your-computer-does-while-you-wait
... but extended to in-proc vs. network calls (assume fast intranet).