Increasing P2P transfers using similar files
A group of researchers from several Universities have come up with a method of file sharing that has the potential to increase P2P speed dramatically. David G. Andersen, assistant professor of computer science at Carnegie Mellon, and Michael Kaminsky of Intel Research Pittsburgh, along with graduate student Himabindu Pucha of Purdue University have come up with a new system for P2P transfer called, Similarity-Enhanced Transfer (SET).
To understand the system, you need to understand how BitTorrent and other distributed file sharing applications work. BitTorrent breaks a file into multiple chunks. When a user goes to download the file, they actually begin downloading each individual chunk. The chunks do not need to be downloaded in order, as each one is assigned a position in the file. As soon as the user has all the chunks, they have the entire file. There are two main advantages to this system. The first one is that the user can download many chunks at a time from many different users. Downloading from multiple users means that the file will be downloaded much faster than if downloaded from a single user. The second advantage, is that as soon as the user has a complete chunk, they can start sharing that piece. So they don't have to wait for the entire file to be available before they start contributing to the availability of the file.
The SET system works much the same way. The difference is that is actively searches for files that are similar to the one being shared, and attempts to download the similar parts as though they were from the same file. This can dramatically increase the availability of a file, which in a protocol like the BitTorrent protocol means that the file will be downloaded much quicker.
News Via PhysOrg: Computer scientists develop P2P system that promises faster music, movie downloads
Published reaserch paper: Exploiting Similarity for Multi-Source Downloads Using File Handprints
Labels: BitTorrent, FileSharing, P2P
0 Comments:
Post a Comment
<< Home