If you discover that a leaf node does not exist for the search key (e.g. there is no leaf for the range you’re seeking), or the leaf node does not contain the desired key, report that the key does not exist, and you’re done. However, you can be a bit more clever by using the same insight from separator key truncation: most database operations are on key ranges, so having fast access to a key prefix can be more important than having access to the entire key. The first row in each node represents the key stored in the node, and the second represents the valid keys that can exist in that node: