In a B+ tree, actual data is stored in

In a hierarchical index, nodes generally serve one of two purposes: guiding a search toward the right location, or holding the actual payload the search is looking for. A well-designed index keeps these two jobs on separate tiers — the guiding nodes stay lightweight (holding only keys and pointers) so more of them fit in a disk block, which increases fan-out and keeps the tree shallow; the payload is confined to one dedicated tier so the rest of the tree never has to carry that extra weight.

Tracing a search through a B+ tree shows exactly where that payload tier sits:

1. The search starts at the root, which compares the search key against its separator keys to choose the correct child pointer — a purely guiding step.

2. It then descends through zero or more internal nodes, each of which again uses separator keys, not payload, to pick the next child pointer downward.

3. The search ends only when it reaches a leaf node; that is the dedicated payload tier described above, so the leaf holds the key together with the actual record or a pointer to the record on disk.

4. Every leaf node is additionally linked to its neighboring leaves through sibling pointers, so once a leaf is reached, a range of records can be scanned in sequence without walking back up the tree.

This is exactly what distinguishes a B+ tree from a plain B-tree: a B-tree may keep data records inside internal nodes as well, whereas a B+ tree deliberately confines all data to the leaf level so every internal node stays small (keys and pointers only, no payload), improving fan-out and reducing tree height. Since the root and internal nodes are defined purely as guiding structures, and 'parent node' is only a relative name for any node that has children (root or internal), none of these levels is where actual data resides in a B+ tree — the leaf level is the sole data-holding tier, exactly the dedicated payload tier the opening principle described. (This routing/payload split describes any tree with more than one level; in the degenerate case of a single-node tree, that lone node acts as both root and leaf together.)