|
Although P2P has proven itself as a viable architectural paradigm for a variety of large-scale distributed applications, P2P is far from reaching its full potential. Peers possess surplus bandwidth, storage and CPU resources, which when aggregated together across all peers worldwide, constitute a huge, untapped resource pool. The success of future P2P applications ultimately depends on convincing users to contribute these resources, which is the challenge of P2P incentive design.
BitTorrent’s tit-for-tat incentive scheme is commendable in that it has been deployed on a massive scale. However, from an economic perspective, tit-for-tat is highly inefficient, since it requires peers to trade synchronously. We believe that the next generation of P2P incentive mechanisms should incorporate some notion of “money,” which users can accumulate for providing services, and later spend for acquiring services.
In this talk I will survey two promising lightweight incentive schemes, allowing P2P users to accumulate money and trade asynchronously. The first scheme, called Network Asynchronous Bilateral Trading (NABT), is a fully decentralized, exploits an underlying social network and pairwise currency among pairs of friends. We have shown that NABT is almost as efficient as a perfect economy, where all users can trade directly with each other. The second paradigm, called Closed P2P Communities, is inspired by current practices in BitTorrent darknets. For closed communities, we will show how mechanism design from game theory can be used to design optimal incentive schemes, and how the notion of entropy get be used to combat colluders.
This is joint work with Yong Liu and Zhengye Liu
|
