Dynamic Multilevel Feedback-Based Searching Strategy in Unstructured Peer-to-Peer Systems

摘要

One of the main challenges of unstructured peer-to-peer (P2P) systems that greatly affects performance is resource searching. The early proposed mechanisms use blind searching, but they have a lot of shortcomings. Informed search strategies have better performance in comparison with blind ones, but they still suffer from low success-rates and long response times due to their inadaptability to dynamic P2P environments where nodes can frequently join and leave the system. To address this problem we propose a new search strategy called Dynamic Multilevel Feedback-based Search strategy (DMFS) in this paper which is greatly adaptable to dynamic P2P systems. DMFS is a new dynamic strategy that pays attentions to nodes heterogeneities and system variable conditions. It deploys important factors to guide resource selection process. These factors include changeable temporal popularity, temporal number of hits, temporal penalty, dynamic ranking, resource heterogeneity, and resource responsiveness status. In addition, DMFS improves system fault-tolerance in case of resources unavailablity. Extensive simulations of DMFS show considerable growth in the resource searching performance by decreasing response time and bandwidth consumption and increasing success-rate.