In various other RDBMS products we can query to see where a given row lives on disk. The idea is that one can monitor and effect the placement of data to be efficient when scanning many rows at once. For example imagine a photo gallery with 100 photos where the application does a logical:

select * from pictures where user_id = 10; / db.pictures.find("user_id":10)

One would want all picture rows in the same block so that only three block fetches are required. Two for index then one for the data payload.

See this pastie for example:

http://pastie.org/936239