Grzegorz Czajkowski discusses on googleblog Google's experiment sorting 10 trillion records of 100 bytes each.

   10 trillion records of 100 bytes =
   10 x 10^12 records of 100 bytes =
   10^13 records of 100 bytes =
   10^13 records of 800 bits

They consider this to be sorting 10^15 bytes and refer to it as a petabyte [sic].

They say it took six hours and two minutes on 4000 computers and 48000 hard drives. On each sort some drives broke. They stored three copies of the resulting files on GFS.

   http://googleblog.blogspot.com/2008/11/sorting-1pb-with-mapreduce.html

Six hours to find the greatest porn on Earth, that's nice ^^