Efficient Flooding in Peer-to-peer Networks

Ai Wu,Xinsong Liu,Kejian Liu

DOI: https://doi.org/10.1109/caidcd.2006.329410

2006-01-01

Abstract:Peer-to-peer (P2P) computing has emerged as a popular model aiming at further utilizing Internet information and resources. Flooding is the basic method of searching in unstructured P2P networks however the blind flooding based search mechanism causes a large volume of unnecessary traffic, and greatly limits the performance of P2P systems. Our study shows that a large amount of unwanted traffic is divinable and can be avoided while searching in P2P networks. Aiming at reducing the volume of unnecessary traffic, an unnecessary message prediction based searching mechanism (UMPS) is proposed. UMPS is a pure distributed scheme, it is based on the distributed neighbor list where neighbors within two hops are stored in. UMPS cuts down unwanted traffic by terminating unnecessary flooding that has been predicted in advance, and the techniques related to distributed neighbor list are also discussed. Simulation results show that more than 60% unnecessary messages can be reduced by UMPS, and the query coverage range is retained at the same time, and more unwanted messages can be reduced if a peer has larger degree, therefore load balancing is achieved also. Simulations also reveal that the effectiveness of UMPS is positive correlation with clustering coefficient of network