A Resource-Based Assessment of the Gnutella File-Sharing Network

This paper reviews the growth behavior of a popular peer-to-peer network. We propose a dynamic hypothesis that the growth, overshoot, and collapse trajectories may be the result of complex causal interactions between inadequate resources, private provision of common goods, free riding, and membership dynamics. We draw parallels with other systems that are well-understood and known to exhibit similar trajectories. Computer experiments confirm that free riding by peers may lead to inadequacy of resources, decline in network performance, high attrition rates, and collapse. However, if freeloading tendencies are not strong, which is usually true in smaller groups, then the P2P system will function without oscillations. An experiment that considers improvements in search algorithms suggests that the reduction of total network traffic may not be sufficient to eliminate system fluctuations in the long run.