A New Structured Peer-to-Peer Architecture Based on Physical Distance

Abstract

Recently, structured P2P (Peer-to-Peer) system has become more and more popular. However, P2P systems' large scale and high dynamics have brought a great challenge to data availability and accessing performance. Redundancy techniques are used to solve these problems. However, in structured P2P systems, due to the consistent hash, the overlay network could not match underlying physical network well. The nodes close to each other in the overlay network may have long distances of physical network. In this paper, we put forward a new P2P architecture that constructs node identifiers and places redundant data according to physical location information. It can provide better load balance and access performance. The node's identifier is divided into four parts, representing the node's state, ISP, city and IP respectively, so the nodes having similar identifiers are close to each other in the physical network. Moreover, a query tree is used to help a node routing queries quickly in the physical network. In addition, we maintain an access list for each file. When a node becomes overloaded, replicas are placed on another node selected in the routing path according to the access list, so subsequent access requests could be met in advance.