BTW

The real complication of this whole approach of "try again another random number if already taken" idea, cause this will slow down considerably after a while if millions of random numbers need to be tried out.

Here are the average times, per 1000 part numbers, for the first 9 millions from the 10 million range:

[1000000]Ave: 0.003626
[2000000]Ave: 0.004084
[3000000]Ave: 0.004650
[4000000]Ave: 0.005443
[5000000]Ave: 0.006492
[6000000]Ave: 0.008238
[7000000]Ave: 0.010892
[8000000]Ave: 0.017433
[9000000]Ave: 0.300273
[download]

So sure, allocating the 9th million takes roughly 100 times as long as the first.

But that still means that unless the OPs company are going to need to allocate 1 million new part numbers in less than 5 minutes, the "slow down" isn't going to be any kind of a problem.

And if they are going to allocate such ridiculously high numbers of new part numbers, that the allocation rate is a limiting factor; then they sure as heck need to be starting with a much larger range, otherwise they will run out in their first hour of trading.

You've yet to learn the difference between knowledge and expertise.

The knowledgeable are aware of theories, concepts and formulae.

The experienced know when to apply them.