IO being short for Input/Output here. These are metrics of the maximum throughput the device can have, in both directions. At a certain point it's simply performing at it's maximum speed and cannot possibly do more work than it is at the time. This would create a bottleneck, causing either errors or wait states, while the requests that are over the maximum threshold will have to be queued, cached or delayed.

The raid array is, of course, a group of disks working together which is why it has a higher throughput but smaller percentage of maximum used when compared to a single drive.