Transferring files over P2P networks can occur much faster than now. Computer researchers from Carnegie Mellon and Purdue universities have created a new protocol, Similarity-Enhanced Transfer (SET), which implements BitTorrent principles at a higher level. Unlike BitTorrent, the SET protocol can split into multiple downloads not only exactly identical files with a matching checksum, but also files that only partially match.
Some files actually do not match at 100%, but at 99.9%. For example, it can be music files with different titles or two versions of a movie with different sound tracks. Using the SET protocol, a gigabyte file is divided into 64,000 fragments of 16 kilobytes - all of them are compared and can be downloaded from different sources.
As you know, the more sources of the torrent - the higher the speed. Depending on the number of sources, using SET, the download speed of a file can grow by 5% and 500%. During the tests, the real increase in MP3 download speed was 71%. On the video trailer, the acceleration was 30% due to the swap of information from other trailers, in which 47% of the content matched.
The source code for the development was shown by researchers at the
4th USENIX Symposium on network designs and implementations . Scientists argue that they do not think about commercial success, but simply want to speed up the exchange of scientific documents on P2P.
')
The authors of the development are Professor David Andersen from Carnegie Mellon and programmer Michael Kaminsky from Intel Research Pittsburgh. They were assisted by a student Himabinda Pucha.
via
EurekAlert