Use of latent semantic indexing for content based searching and routing of mobile agents on P2P network

The peer-to-peer (P2P) system has a number of nodes that are connected to each other in an unstructured or a structured overlay network. One of the most important problems in a P2P system is locating of resources that are shared by various nodes. Techniques such as Flooding and Distributed Hash-Table (DHT) have been proposed to locate resources shared by various nodes. Flooding suffers from saturation as number of nodes increase, while DHT cannot handle multiple keys to define and search a resource. Various further research works including multi agent systems (MAS) have been pursued that take unstructured or structured networks as a backbone and hence inherently suffer from problems. We present the solution that is more efficient and effective for discovering shared resources on a network that is influenced by content shared by nodes. Our solution presents use of multiple agents that manage the shared information on a node and a mobile agent called Reconnaissance Agent (RA) that is responsible for querying various nodes. To reduce the search load on nodes that have unrelated content, an efficient migration route is proposed for RA that is based on cosine similarity of content shared by nodes and user query. Results show reduction in search load and traffic due to communication, and increase in recall value for locating of resources defined by multiple keys using RA that are logically similar to user query. Furthermore, the results indicate that by use of our technique the relevance of search results is higher; that is obtained by minimal traffic generation/communication and hops made by RA.

Available electronically through Sheppard Library, Middlesex University.