The traditional way to handle it is to have a central database the assigns numbers and increments a counter. The problem with is the central database may be inaccessible. Further the central database becomes a crucial bottleneck.

I suggest a partially decentralised variant in my Unique Number Server Student Project.

Another way to do it is to use large pseudorandom numbers and count on the improbability of two numbers being picked the same. This technique is fully decentralised approach. No communication between number assigners is needed. This is how java.util.UUID works.

Learning More