Shrinking a database file is performed by manipulating pages within the file such that all empty pages reside at the end of the file. Non-empty pages from the end of the file are written to empty pages at the start of the file. Repeating this process, the correct number of times, results in all of the empty pages residing at the end of the file. Once this is achieved, the file is truncated such that the space consumed by all empty and discardable pages is returned to the file system. Truncating the file involves traversing the file backwards from the end of the file until the first non-empty page is found. The file is truncated after the last non-empty page.