Clustering an unstructured P2P networks using a termite hill building model

Abstract

Super-peer (SP) architecture is proposed to improve the quality of service (QoS) of peer to peer (P2P) networks. P2P networks is divided into sets of homogeneous sub-groups representing the number of SPs. Designing SP networks for file sharing has several issues like the specifying best number of SPs, selection of SPs, and suitable ordinary peers for each SP. In this paper, we propose a simple method to achieve self-organization of peers in dynamic environment to enhance QoS. Termite hill building model is used for clustering an unstructured P2P network by employing Jaccard measure to compute peers’ interest similarity. This method consists of four steps which are initialization, separation, colony building, and post processing. Both the separation and colony building steps are the backbone of the method. The experimental results on a simulated network with 10000 nodes show about 99% as accuracy.