Perhaps the first example of seamless user-contributed content on the internet was Napster, the music sharing software that reached over 25 million users over 3 years before it was shut down in 2002.
Napster paved the way for peer-to-peer file sharing programs like Kazaa and LimeWire as well as user-contributed content sites like YouTube and MySpace.
My own research in peer-to-peer networks arose out of using and observing these networks, and is described by the following 4 themes:
Reputation - One of the biggest problems in P2P Networks today is inauthentic file attacks. It has been suggested that the future of file-sharing P2P Networks depends on the ability to know the reputation of file sources in the network. However, reputation management in open, anonymous, distributed networks presents a number of challenges. EigenTrust is an algorithm that manages reputations in such networks.
Simulation - Testing algorithms on file-sharing P2P networks is often done by simulations, since deploying algorithms on real P2P networks is often impossible. However, many P2P algorithms are sensitive to network and traffic models that are used in simulations. Therefore, to accurately test P2P algorithms, we require a simulator that resembles real-world file-sharing P2P networks as closely as possible. The Query-Cycle Simulator is a file-sharing P2P network simulator that is standardized, extensible, and modeled after measurements in real-world file-sharing P2P networks. We are making the code available so that research groups can test their algorithms on a standardized P2P network simulator
Topologies - We present a peer-level protocol for forming adaptive topologies for file-sharing P2P networks that is based on the idea a peer should directly connect to those peers from which it is most likely to download satisfactory files. We show that the resulting topology is more efficient than standard Gnutella topologies. Futhermore, we show that the resulting topology has the added benefits of increased resistance to certain types of attacks, intrinsic rewards for active peers, and intrinsic punishments for malicious peers and freeriders.
Market Mechanisms - The operation of large-scale competitive P2P systems are threatened by the non-cooperation problem, where peers do not forward queries to potential competitors.
While non-cooperation is not a problem in current P2P free file-sharing systems, it is likely to be a problem in such P2P systems as pay-per-transaction file-sharing systems, P2P auctions, and P2P service discovery systems, where peers are in competition with each other to provide services. Here, we motivate why non-cooperation is likely to be a problem in these types of networks and present an economic protocol to address this problem. This protocol, called the RTR protocol, is based on the buying and selling of the right-to-respond (RTR) to each query in the system.
By Sepandar D. Kamvar, Mario T. Schlosser, and Hector Garcia-Molina
By Mario T. Schlosser, Tyson E. Condie, and Sepandar D. Kamvar
By Sepandar D. Kamvar, Mario T. Schlosser, and Hector Garcia-Molina
By Sepandar D. Kamvar, Mario T. Schlosser, and Hector Garcia-Molina
By Beverly Yang, Sepandar D. Kamvar, and Hector Garcia-Molina
Sepandar D. Kamvar
Sepandar D. Kamvar