The relevance of introducing optical interconnects (OI's) in monoprocessors
and multiprocessors is studied from an architectural point of view. We sho
w that perhaps the major explanation for why optical technologies have near
ly been unable to penetrate into computers is that OI's generally do not sh
orten the memory-access time, which is the most critical issue for today's
stored-program machines. In monoprocessors the memory-access time is domina
ted by the electronic latency of the memory itself. Thus implementing OI's
inside the memory hierarchy without changing the memory architecture cannot
dramatically improve the global performance. In strongly coupled multiproc
essors the node-bypass latency dominates. Therefore the higher the connecti
vity (possibly with optics), the shorter the path to another node, but the
more expensive the network and the more complex the structure of electronic
nodes. This relation leaves the choice of the the best network open in ter
ms of simplicity and latency reduction. The bottlenecks resulting from and
the benefits of implementing OI's are discussed with respect to symmetric m
ultiprocessors, rings,: and distributed shared-memory supercomputers. (C) 2
000 Optical Society of America.