DPAS: A dynamic popularity-aware search mechanism for unstructured P2P systems

Abstract

One of the pivotal challenges of unstructured Peer-to-Peer (P2P) systems is resource discovery. Search mechanisms generally utilize blind, or informed search strategies wherein nodes locally store metadata to quicken resource discovery time compared to blind search mechanisms. Dynamic behavior of P2P systems profoundly affects the performance of any deployed resource-searching mechanism and that of the whole system in turn. Therefore, efficient search mechanisms should be adaptable to the dynamic nature of P2P systems whose nodes frequently join and leave the system. Nonetheless, existing informed search mechanisms have failed to accord with dynamicity of P2P systems properly, thereby becoming inefficient. To address this issue, we put forth a new resource-searching mechanism called Dynamic Popularity-Aware Search mechanism (DPAS). Our mechanism estimates the dynamic responsiveness states of candidate nodes to direct search selection process by exploiting temporal number of hits, temporal penalty, and node heterogeneity. Besides, it controls the search scope at each step by estimating both the dynamic temporal popularity of resources and recently obtained results. It thus considers at each step of the search decision-making process to conform itself with the dynamics of P2P systems. Extensive experiments have demonstrated that DPAS has enhanced performance in comparison to other pertinent search mechanisms by virtue of an upsurge in the success-rate and decrease in the response time and bandwidth consumption.