If you are referring to the b-tree in how data is stored in both the table and the index, adding a clustered index keyed off of an incrementally generated identifier (think bigint) will do the trick. This will physically store the data sequentially in the underlying pages and extents. If your tables do not have a clustered index they will be stored as a heap and there will be no order to the storage of the data. If you are referring to indexes and the b-tree that is associated with index storage, I would recommend implementing a fill factor to allow for empty space to accommodate new rows to be added. This will not eliminate page splits but can help to reduce them.