Without a unique name, the scheduler (and anything else, really) will
consider each of these nova-compute processes as being one and the same,
so only one entry in the services table, only one hypervisor seen by the
scheduler, etc. Assigning unique names lets us simulate an arbitrary
amount of nova-computes which is very handy for benchmarking the
scheduler.
Change-Id: Ie03aad81bd2a8e73b876a9eae934bc00bf2f71e9