I think the programming is the tough part. Something like cracking password
hashes does very well on parallelized machine. Just hand of chunks of work to
each, and your all set. But a lot of problems are much more complex. I
believe with some problems the speed increases with the root of the number of
cpu's / machines rather than being linear.
VMS lends itself to clustering, but I doubt it's used on supercomputers. You
could have SIMH running on linux then get a copy of VMS/VAX and run it on each.
The hobbyist license would be $30. Then you could cluster the VMS machines.
Just an idea. I'm going to set up some emulated VAXEN at some point.
The O'reilly book covers linux clusters. Don't know about windows clusters or
how that is done.
Michael
On windows it works like this:
win2kpro- 2 cpu's max
win2k server- 4 cpu's max
win2k advance server- anything over 4
One of the links I posted is to a site that has a open source program
for running windows clusters and task allocation.
Just do a dogpile or google for windows clusters and you'll find 100's
of sites that show how it's been done.
Not saying it's simple just that with a bit of sweat it can be done.
I've got several 1gig amd machines sitting in the back shop waiting
for the day when I can get around to building one for myself.
I do msotly video rendering and it'd speed up my business greatly and
with near zero outlay. Which is the main point of cluster farms, using
old gear to get near gigaflop performance.
