Use of Rowids

Every data row in a nonfragmented table is uniquely identified by an unchanging rowid. When you create an index for a nonfragmented table, the rowid is stored in the index pages associated with the table to which the data row belongs. When the database server requires a data row, it searches the index to find the key value and uses the corresponding rowid to locate the requested row. If the table is not indexed, the database server might sequentially read all the rows in the table.

Eventually, a row might outgrow its original storage location. If this occurs, a forward pointer to the new location of the data row is left at the position defined by the rowid. The forward pointer is itself a rowid that defines the page and the location on the page where the data row is now stored.