Design and Implementation of a Novel Peer-To-Peer Information Retrieval Framework

Abstract

Peer-to-Peer (P2P) Information Retrieval framework consists of a peer to peer network of nodes, which voluntarily agree to share their resources by joining the network. While joining these nodes construct the active peer list. Each peer maintains a B+ tree containing IP hash values. The files are distributed over the peer to peer network based on the keywords. The files are initially uploaded into the target node based on the closest match between the hash values of the IP address of the node and the keywords used to index the file. While searching, the target node is identified by finding the closest match between the hash value of the keyword and the IP address hash from the B+ tree stored in the peers. After identifying the target node, the references to desired document is retrieved by searching a B+ tree indexed using keywords. The proposed framework uses Hadoop cluster to extract keywords from the files to be uploaded in the desired target node. Hadoop’s MapReduce programming paradigm reduces the time for keyword extraction. As the framework maintains a B+ tree in the peers, it further reduces the search time and improves network bandwidth.