Incentives - Peer-to-Peer Computing: Applications, Architecture, Protocols, and Challenges - page 86

the penalty mechanism is effective in deterring free-riders when the penalty is
higher than the contribution cost. In quantitative terms, the condition is that
p > 1/α. Finally, another interesting conclusion is that for a su ciently heavy
penalty, no social cost is incurred because every user will contribute (i.e.,
choose not to be a free-rider) so that optimal system performance is achieved.
In particular, to deal with the whitewashing problem, the analysis suggests
that every new user is imposed a fixed penalty. Essentially, this is similar to
the case in the eBay system where every new user has a zero reputation and
thus, will less likely be selected by other users in commercial transactions.
However, this is in sharp contrast to the adaptive stranger policies suggested
also by Feldman et al. in another study [Feldman et al., 2004a] that we have
described earlier.

5.2.1.7 Game Theoretic Modeling

Ranganathan et al. [Ranganathan et al., 2003] proposed and evaluated
three schemes induced by the Multi-Person Prisoner's Dilemma (MPD) [Os-
borne, 2004, Schelling, 1978]. The basic Prisoner's Dilemma game models the
situation where two competitors are both better off if they cooperate than
when they do not. However, without communication, the unfortunate stable
state is that both competitors would choose not to cooperate. An MPD is
a generalization of the basic PD. Specifically, the key features of the MPD
framework can be briefly summarized as follows:
•The MPD game is symmetric in that each of n players has the same
actions, payoffs, and preferences.

•Any player's payoff is higher if other players choose some particular
actions (e.g., “quiet” instead of “fink”).
The MPD framework is used for modeling P2P file sharing as follows. There
are n users in the system, each of which has a distinct file that can be either
shared or kept only to the owner. The system is homogeneous in that all files
have the same size and same degree of popularity. Now, the potential benefit
gained by each user is the access of other users' files. The cost involved is the
bandwidth used for serving other users' requests. With this simple model, it
can be shown that the system has a unique Nash equilibrium in which no
user wants to share. Obviously, this equilibrium is sub-optimal (both at the
individual level and at a system-wide level) in that each user could obtain a
higher payoff (i.e., a higher value of net benefit) if all users choose to share
their files.
Motivated by the MPD modeling, Ranganathan et al. proposed three in-
centive schemes:

•Token Exchange: This is a payment-based scheme because each file
consumer has to give a token to the file owner in the sharing process.
Each user is given the same number of tokens initially and each file has
the same fixed price.