The major difference between ISAM and B-tree is that the B-tree index grows as the table grows. If you added these five new employees to the ISAM employee table, keyed on name: Zanadu, Zentura, Zilla, Zorro, Zumu, these names are put on the last page of the ISAM table. Because they do not all fit on the last page, they are put onto an overflow page attached to the last page.

If you added these five new employees to the B-tree table, you add the new names to the appropriate leaf page (Page 4, in this case) and their records go on the associated data page for leaf Page 4. Because the associated data page fills up, a new associated data page is assigned to Page 4. If the leaf page is full, and cannot hold all five names, the leaf page splits into two leaf pages, and a reference to the new leaf page in the index is added. If the index page can no longer hold a reference to another leaf page, the index is split as well.

Splitting occurs fairly frequently while the table is small and growing. As the table gets larger, splitting occurs less frequently (unless a sequential key is used) and usually only in the leaf or lowest index level.

Repeated inserts into the right-most leaf of a B-tree table create empty leaf pages rather than half-full ones. This improves insert and retrieval performance, and increases disk space efficiency.