A Practical Method for Detecting Community Structures in Decentralized and Unstructured P2P Networks

In decentralized and unstructured P2P networks like Gnutella, there is no coupling between topology and data location. Nodes can leave the network arbitrarily, which greatly affects their neighbors. We should determine in advance how many and which nodes will be affected when some nodes leave the network due to unstable Internet environment. In view of this challenge, we model the problem as detecting community structures, regarding the leaving nodes as community cores. Just as the flooding data queries in decentralized and unstructured P2P networks, the algorithm also works in a flooding way. It starts from the community cores and assign their neighbors to each corresponding community. Specifically, we focus on decentralized and unstructured P2P networks only. At last, a case study is presented for validating the method.