Technology

An Internet test was taken this week and it showed that “intelligent” peer-to-peer traffic can dramatically lower network utilization and speed up downloads for any subscribers. Verizon participated in the study – which was headed by Yale University – will be releasing all of the details this week at the Distributed Computing Industry Association’s P2P Market Conference in New York City.

Using network topology data from Verizon and Telefonica, Yale University tested a software enhancement to the peer-to-peer protocol that it developed with software developer Pando Networks. What the researchers discovered was that when using the so-called P4P software they were able to reduce the impact of peer-to-peer traffic on Verizon’s network by more than 50 percent. This is significant because peer-to-peer traffic makes up roughly half of all traffic traveling over Verizon’s network.

The P2P protocol, which is used to distribute large data files, works by requesting pieces of a single file from different hosts all over the Internet. The technology has become popular for distributing high-definition video. But applications that use P2P eat up a lot of bandwidth, which some service providers say is a problem. Cable operator Comcast has slowed down certain kinds of peer-to-peer traffic in an effort to manage its network. And Time Warner Cable is experimenting with a tiered usage model to deter people from sharing P2P files.

Traditionally, the P2P protocol has requested bits and pieces of content randomly, without considering the physical location of the data. This often results in some pieces of the content traveling over long distances across the network. For example, a user in New Jersey downloading a movie might get some bits of the file from New York and others from China or California. The P4P software enhancements add intelligence to this process so that the bits are served from local hosts.

Douglas Pasko is Verizon senior technologist and co-chair of the P4P Working Group, which was formed by Verizon, Pando Networks, and the university to develop P4P. He said that when the P4P software was used on the Verizon network it found that 58 percent of its peer-to-peer network traffic stayed local. Using regular P2P technology, only 6 percent of the traffic stayed local.