Game Theory and Peer-to-Peer Sharing
File sharing on the Internet is the process of obtaining access to digitally stored information online.Nowadays, one way of sharing files through the internet is known as the Peer-to-peer file sharing.Peer-to-peer(P2P) communication is provided through the sharing of widely distributed resource in the Internet, there is no specific server needed and individuals collect the information from multiple servers and other peers in the network . In typical P2P protocols files are broken into pieces and are recollected by each individual through the communication . P2P communication scales well with a large number of clients and therefore decreases the amount of time needed for file transferring significantly.
In Peer-to-Peer networks, game theory is widely applied to help regulate behaviors of participants, build sharing incentives for them and maximize the efficiency of the network. The main built-in reasons for people to share resources, as regulated by the game itself, include altruism, need for new resources and fear of punishment. Therefore, for the network, different incentives,or sharing strategies of the “players” are constructed to achieve the optimal network outcome. The common used incentives include: inherent altruism when the sharing cost is low; monetary schemes for which users pay when they obtain resources and get paid when they share them; reciprocity strategy for which participants build reputation by his/her own sharing behaviors, and this reputation will lead to reward or punishment to his/her future sharing experience.
One important issue with Peer-to-Peer network is the free-riding problem. Some peers tend to only download instead of uploading resources to other peers. Better sharing strategies can be used to prevent the number of free riders from being unreasonably high. In the paper of Teemu Typpi, he introduces an sharing incentive presented in “Game theory as a tool to strategize as well as predict nodes behavior in peer-to-peer networks” by R.Gupta and A.K. Somani. In this network model, users is more likely to obtain service if his/her reputation is high and his/her reputation will increase if he/she accepts a service request. Gupta and Somani modeled it as infinitely repeated games and the user or player can choose either to serve or not serve others. It proved that the probability of sharing in the Nash equilibrium of the game is always larger than 0, which means that being a free rider is simply a bad strategy. The author also showed that the system performed better if more past actions were taken into consideration. It is suggested that by modifying sharing strategies, the network can have a better performance.The current main Peer-to-Peer networks commonly adopt the reciprocity way of establishing the user incentives, some work has been done about using more complicated monetary sharing strategies but it still needs more exploration to help network to achieve its maximum.
Source:http://www.cse.tkk.fi/en/publications/B/5/papers/Typpi_final.pdf
gbz1991