I've done this before. You don't really mean process migration, you mean
remote execution unless I misunderstood. Migration is something that you
do to an existing process, remote execution is something you can do to
a new process. The former is (a) a bad idea, and (b) extremely difficult and
invasive to the kernel to fully implement. Fortunately, remote execution
is extremely useful and quite easy to implement. In fact, I've just started
rewriting a version of this I had running in Sun about 6 years ago.
I had a version of make(1) that did all this stuff, I got a 45 minute
to 3.5 minute speedup on a cluster 8 SS2's plus a 4/470 file server.
Not bad, eh?
I'm working on a new version of this same idea in Linux. I'm working
on this with Don Becker & Dan Ridge of Beowulf fame and I need to rope
Tridge (of AP1000 fame) into the project as well.
I think it would be fun to keep in touch on this topic and those of you
who are really interested in this sort of thing as a job should contact me
directly. I have openings for this sort of work right now.
--lm