An approach based on locality-awareness and interest-focusing for web content distribution in P2P networks

A growing number of collaboration application are being built on top of Peer-to-Peer (P2P) networks which provide scalability and support dynamic behavior. However, the P2P networks is built over physical network and each node is logically connected to its neighbor, which lead to data transfers over long distance between nodes as well as to cause high latency and communication overheads. In this paper we propose an approach that exploits the locality information and content information between peers to distribute content rapidly in P2P networks. The idea is that peers keep the web content they accessed and serve it to other peers that are close to them in locality, then the peers build a structured which similar to Kademlia content overlay identified by locality and website. When a new client requests some content from a website, an improved-DHT quickly find an index-peer which ensure the nearest neighborhood peer that has the content available in the locality. According to the heterogeneity of peers, we present a formula to select an optimal peer from the content peers as the index-peer to charge them. Combine with the physical network layout, the website server and index-peers establish a tree structure to change information between them. After a complete theoretical analysis, the new complex structure takes advantage of the actual physical proximity of peers in the P2P networks to reduce network latency, and improves the speed of site content distribution and the site access.