Analyzing Network Coverage in Unstructured Peer-to-Peer Networks: A Complex Network Approach

In this paper, we apply the theory predicting neighbor distribution of arbitrary random graphs to analyze the network coverage of the peers in unstructured peer-to-peer(p2p) networks that use TTLbased flooding mechanism for search and query. However, we find that for many cases, the theory cannot be directly applied to obtain correct estimate of network coverage due to the presence of certain types of edges that we refer as cross and back edges. It is also observed that the presence of cross and back edges in the p2p networks reduce the coverage of the peers and also generates large number of redundant messages, thus wasting precious bandwidth. We refine the theory and develop a model to estimate the network coverage of the peers in the presence of cross and back edges. We simulate our model for different networks with various degree distribution properties. The results indicate that our models provide good estimates of second neighbor and network coverage distribution. We perform a case study of the Gnutella networks to analyze the effects cross and back edges on network coverage and message complexity in these networks. Based on our study, we propose a new bootstrapping algorithm for Gnutella networks named HPC5 that substantially improves the network coverage and message complexity. The results have been validated using simulations.