Carnegie Mellon speeds up P2P transfer

Posted by Humphrey Cheung

Cambridge (MA) - Enhanced file searches could help speed peer to peer file transfers according to new research done by the Carnegie Mellon University and Intel.  Researchers claims their system of searching for similar data files – instead of just identical files like traditional P2P programs – could garner a five-fold increase in transfer rates.

David Andersen, assistant professor of computer science at Carnegie Mellon and Michael Kaminsky of Intel Research have dubbed their system Similarity-Enchanced Transfer or SET.  SET chops up files into 16 kilobyte chunks which can then be searched for through P2P networks.

Traditional P2P networks transfer and recombine identical data chunks from a larger file.  Both the chunks and the larger file must be identical which, according to Andersen and Kaminsky, reduces the numbers of available sources to download.  Downloading similar or identical chunks from any available file would presumably increase transfer rates.

The researchers say many music files are 99% similar and only differ in the artist and title headers.  Software packages are also highly similar.

According to the researchers, SET could increase download speeds of MP3 music files by 71%.  Large movies could be downloaded up to 30% faster while smaller movie trailers could be seen up to 47% faster.