Date: Around 1995
What is the motivation for architectures like an n-cube, n-torus, or a
n-sphere, in massively parallel machines. It seems that the possibilities
of architectures of this kind are limitless. I really do not know anything
about computer architecture, but do not keep your response simple! Where
would I find information about computer architecture?
The motivation is basically to try to allow even quite far away processors
to communicate quickly with one another. Nobody really knows the best way
to do this (as far as I know) and I think the motivation is really just
people experimenting with different ideas to see what works. Like you say,
the possibilities seem limitless.
The constraints are: (1) you do not want any processor to be more than X steps away from any other, and (2) you are limited to Y connections into and out of each processor. Right now it looks like tree and 2- or 3-D mesh structures are winning out, but who knows.
You might check the USENET newsgroup computer parallel for information, or
check some of the journals from IEEE, ACM or SIAM - there are probably some
that focus on parallel machine topology, though I do not know the names.
Click here to return to the Computer Science Archives
Update: June 2012