Using two-tier bitwise interest oriented QRP data structure for high clustering through interest similarity in SWB P2P systems

Lim, Tong Ming * and Chong, Yong Liang * and Lee, Angela Siew Hoong * (2012) Using two-tier bitwise interest oriented QRP data structure for high clustering through interest similarity in SWB P2P systems. In: International Conference on Information Technology: New Generations (ITNG)(9th) , 16-18 April 2012, Las Vegas, NV. (Submitted)

Preview

Text Lim Tong Ming - Using Two-Tier Bitwise Interest Oriented QRP Data Structure for High Clustering.pdf - Submitted Version Download (290kB) | Preview

Abstract

The concept of “Small-World” paradigm has been used by many peer-to-peer (P2P) systems to achieve high clustering and low number of hops to arrive to its desired target. This approach intends to improve performance of P2P systems. The Small world concept requires that the architecture of a P2P system to achieve high cluster coefficient and low average hop between any two randomly chosen peers. In this research, we propose an enhanced hierarchical overlay network by incorporating the concept of Small world into the base hierarchical architecture. This research adopts the Query Routing Protocol (QRP) data structure of a hierarchical P2P network by storing the interest information of files in the leaf peers in the network. The QRP in the leaf peers will be aggregated and propagated to the SuperPeer so that interest information could be used to form short-range, medium-range and long-range links with other SuperPeers to achieve low average hop. As peers join and leave the network, a proposed rewiring protocol is used to ensure peers are clustered by interest to form high clustering coefficient so that search activities are yielding higher relevance results in a more predictable fashion. We simulated the porposed small world P2P using test scenarios to evaluate recall rate of the small world P2P system. Our simulated results evaluate and benchmark our proposed Small-World Bee (SWB) overlay network with iCluster [3], Firework Query Model [5] and Limewire [1] to assess performance obtained and goals achieved in the research work so that future research directions could be planned.

Actions (login required)

View Item