I re-ran some of the Onion Routing benchmarks with only a single
CPU. One of the CPUs was turned offed using psradm. While Michael
pointed out from the man pages that this does not ensure that
the processor will not be used, the man pages does say that
the CPU will be given little or no work. Using this command
was alot easier than pulling out the CPU and also avoided the risk
of damaging the CPU and causing the 2 processor configuration to be
permanently configured as a single processor. 

After executing the psradm command, the system did seem noticeably
slower. I ran the benchmark for the non-crypto case and the throughput
was about 1/2 of that for the 2 CPU case. Hmmm, this seemed mighty fishy !
I ran the crypto-norsa case, again got about 1/2 the throughput again !
Things were looking highly suspicious at this point ! As a sanity check,
I ran my client/server without the Onion Router, the throughput was reduced
by a 1/3. O.K. now the problem is starting to dawn on me.

What I think you wanted to do David, was limit the Onion Routing *processes*
to a single CPU, not the whole system. By turning off a CPU, we changed
the whole complexion of the performance of the OS ! Thus we cannot 
attribute performance difference solely due to the lack of parallelism
since we are no longer comparing like objects because now the whole system
has changed. There is no way that we can constrain a set of processes to
run on a single CPU, so I don't think we can easily study the parallel/
non-parallel problem very easily.