Remaining questions
Are large synchronous computers actually of any use? For the majority of
applications large parallel computers don't need to exhibit large
scale synchronous behavior, as the operation of their components need not
be completely deterministic to be useful. Other factors, such as the
susceptibility of computers with large component counts to point failures,
also make building large deterministic computers problematical.
If it is not absolutely essential, using a computer which must be
deterministic should be avoided wherever possible. However it seems likely
that humanity will continue to find uses for deterministic operation.
The demonstration here is designed to illustrate the idea that the
possibility of generating global synchronous behavior through local
interactions is of relevance to the clocking of large computers.
Much work needs to be done before it can be said that the concept is
properly proven.
The simulation here makes no attempt to simulate a system composed of
a network of analogue components with varying response times, in order to
illustrate how this might exhibit global synchrony. More complex and
realistic simulations are probably the most appropriate next stage in
the development of the idea.
The mixed strategy presented here was located by a trial-and-error process.
No claim is made about how well it solves the target problem. Exactly what
type of firefly system is most appropriate as the basis of a large
deterministic computer is still very much an open question.
Lastly, there are questions of power consumption and heat dissipation to be
addressed. The 'fast' automata presented here is reversible everywhere,
except at its edges. The other automata are not fully reversible
- and it appears that if they are to retain the property of being resilient
in the face of errors, they cannot be made perfectly reversible.
Making a firefly-system with very low power consumption and heat
dissipation figures should be possible, however.