文摘
Peer-to-Peer (P2P) networks are collaborative distributed systems which are not dependent only upon external servers to provide the required resources to users. By downloading files from the servers as well as uploading them to the others, the peers mitigate the server responsibilities. One of the most important metrics to design a peer-to-peer network as an overlay network, is fairness. To get better levels of fairness, detecting and preventing free-riding behaviors are necessities. Most works in the literature do not have the capability of totally preventing free-riding, rather they just make it slightly harder. However, there are few approaches which can successfully prevent free-riding at the cost of poor efficiency and slow downloading, as peers select their neighbors randomly. We propose a more intelligent approach that prevents free-riding completely and provides better overlay formations to improve the performance metrics. The key idea is to share knowledge among peers to learn about their environment. Based on this knowledge, the peers dynamically change the overlay network in a way that they select the best neighbors for resource reciprocation. Empirical results, obtained through simulations, show our approach outperform the other existing methods on both stopping free-riding and increasing the P2P network performance.