I built a file storage web system several years ago (very popular when free, but when the free beta ended not many people wanted to pay). We computed the hash of each uploaded file, and only stored one copy of the file based on the hash. The database based filesystem simply recorded the hash of the file as basically a pointer to the actual file. This allowed multiple people to have the same file, with different file names (stored in the db), but only actually store the file once. Very simple.