Multiple CPU's and interrupt latency / "determinism"

[Len Gross]

Is there a way to use multiple CPUs so that the interrupt latency of
the system is bounded and presumably smaller than a single CPU system?
 That is, can I have a process isolated so that for all practical
purposes it is real-time?

My particular issue is needing to be able to transmit network packets
at very precise times.

Thanks in advance.