hashing is key-to-address translation. This means the value of a key is transformed into a disk address by means of an algorithm, usually a relative block and anchor point within the block. It's closely related to statistical probability as to how well the algorithms work.

It sounds fancy but these algorithms are usually quite simple and use division and remainder techniques. Any good book on database systems will have information on these techniques.

Interesting to note that these approaches are called "Monte Carlo Techniques" because the behavior of the hashing or randomizing algorithms can be simulated by a roulette wheel where the slots represent the blocks and the balls represent the records (on this roulette wheel there are many balls not just one).