The math gets a little less obvious with something like RAID-5 since there isn't a designated
parity drive. The parity data gets distributed among all the drives. But for RAID-5 you'll always need one additional
drive to do an array.
Minimum number is three drives. Your data space is the capacity of all
the drives in the array minus the capacity of one
drive in the array. (Note: All the drives in a pure RAID-5 array must have identical
Assuming 1TB drives, a three drive array would have a 2TB data space (i.e. 3TB-1TB). So 33% of the total capacity is 'lost' to parity. But a five drive array would have a 4TB data space (i.e 5TB-1TB) with only 20% total capacity lost to parity. And that percentage decreases with each drive added.
So as you can see, RAID-5 is least
economical with only three drives - but it becomes increasingly
economic with each drive added after that - up to the capacity of the RAID controller.
The tradeoff is that a standard RAID-5 array can only tolerate one
drive failing at a time without experiencing data loss. As skwire
pointed out earlier, the general rule is you can have as many drives simultaneously
fail as you have parity drives. And in the case of RAID-5, you basically only have one "drive" for parity even though it isn't a specific drive.
In practice, it's a little more complicated than that. But not much.