Unless the machine has multiple interface cards, there is very little if any benefit in using multiple threads or processes, or an event driven architecture, as regardless of the presence of multiple cpus, all the outbound packets will be serialised at the interface card.

He might as well just run a tight loop issuing requests as fast as the processor/interface can send them.

As UDP is connectionless, at most, he might benefit from two threads/processes per cpu. One to send as fast as possible and another to receive replies assuming he is going to bother.