Gpop: A Global File Popularity Measurement for Unstructured P2P Networks

Gpop: A Global File Popularity Measurement for Unstructured P2P Networks

Manel Seddiki (Computer Science Department, University of Sciences and Technology Houari Boumediene, Algiers, Algeria) and Mahfoud Benchaïba (Computer Science Department, University of Sciences and Technology Houari Boumediene, Algiers, Algeria)
Copyright: © 2015 |Pages: 14
DOI: 10.4018/IJDST.2015070104
OnDemand PDF Download:
$30.00
List Price: $37.50

Abstract

Unstructured overlays such as P2P networks and social networks stimulate many research areas. This kind of overlays is composed of a set of self-manageable entities which share objects between them in a spontaneous way. Getting a global knowledge such as popularity of shared objects or reputation of the entity is a challenging task because in such overlays, entities have only partial knowledge about the overlay state. In this paper, the authors focus on the file popularity measurement because this parameter can be efficiently used to improve object replication and object search performances. Some research works are proposed to measure this parameter, but these measurements are only based on local knowledge of peers. The authors propose Gpop, a global file popularity measurement for unstructured P2P networks which considers both local knowledge of the peer and knowledge of the other peers participating in the network to gain a global-like knowledge. Simulation results reinforce the authors' theoretical propositions and show that our measurement is closer to the real file popularity.
Article Preview

Introduction

The Peer-To-Peer (or P2P) concept has been proposed to address the centralization problem of client / server systems. The basic idea of P2P is to link entities called peers in order to exchange information without using any intermediate server. A P2P network is a distributed system of interconnected peers which are both clients and servers. The P2P paradigm was firstly used for file-sharing applications such as in (Napster) and Gnutella (Matei et al. 2002) which allow users to search for, share and download files but the concept was extended to handle other objects such as CPU, bandwidth and services and so on.

The first P2P network which has been appeared is Napster. It uses a server to index all information about peers and their shared files. If a peer wants to search for a file, it sends a request to the server which connects it with peers storing this file then, the peer directly downloads the searched file from the selected peer. Indexing all files in the server facilitates search procedure and improves the search latency, but if this server fails, the whole system fails. Gnutella succeeded to Napster and erased the centralization issue. It introduces the decentralized unstructured P2P network but in this latter, each peer is blind and has no knowledge about other participating peers and their shared files. A peer wishing to search for a shared file broadcasts its request to all its neighbors, which do the same with their neighbors until the file is found or the Time To Live (TTL) expires. This technique is called flooding (Matei et al. 2002) but its main drawback is the high overhead that in turn causes scalability issues. Many alternatives to flooding have been proposed to reduce overhead issue and make file search more efficient, such as using probability based on previous search results (Gaeta& Sereno 2002),(Margariti 2011) using progressive TTL called Expending Ring Search (Lv et al. 2002) or using Random walk technique (Gkantsidis & Saberi 2004).

P2P networks have stimulated many research areas such as objects’ search (Sharifkhani & Pakravan 2013), objects’ replication (Thampi & Sekaran 2009), incentive to cooperation, trust propagation in social networks and so on. These research works in these areas are based on the estimation of some global parameters such as object’s availability (Gao et al. 2013), (Ma et al. 2013), user’s trust (Adali et al. 2010), (DuBois et al. 2011), (Kim & Song 2011),(Sherchan et al 2013),(Huang et al. 2013), peer’s reputation (Dellarocas 2006),(Gupta et al. 2003),(Jurca & Faltings 2003),(Kamvar et al. 2003), (Koutrouli & Tsalgatidou 2013),(Marti & Garcia-Molina 2004), (Yu et al. 2004),(Walsh & Sirer 2006) and object’s popularity (Thampi & Sekaran 2009), (Mohammadi et al. 2010), (Kangasharju et al. 2002),(Meroufel & Belalem 2012). In unstructured networks, these global-parameter estimations are challenging tasks because the peer cannot know the overall state of the network. Usually, the peer acquires information on these parameters based on its local knowledge and those derived from messages passed through it. We focus on this issue and we propose a global-like measurement which allows peers to gain global knowledge about the network and thus, estimate global-like parameters. We choose object popularity as a global parameter because this measurement can be efficiently used in object search, replication and retrieval strategies which are a big research area in unstructured P2P networks to improve network performances. We have already proposed a file popularity measurement in a first contribution (Seddiki & Benchaiba 2013). This measurement is based both on peer’s local estimation and estimations of other peers participating in the network. Indeed, considering the other peers allows having a global-like popularity which is more realistic. In this paper, we extend our initial work by adding: i) Additional details about the proposal, ii) Additional simulation results which confirm and reinforce our equations, iii) An improvement based on simulation results for our global measurement.

Complete Article List

Search this Journal:
Reset
Open Access Articles
Volume 9: 4 Issues (2018): 1 Released, 3 Forthcoming
Volume 8: 4 Issues (2017)
Volume 7: 4 Issues (2016)
Volume 6: 4 Issues (2015)
Volume 5: 4 Issues (2014)
Volume 4: 4 Issues (2013)
Volume 3: 4 Issues (2012)
Volume 2: 4 Issues (2011)
Volume 1: 4 Issues (2010)
View Complete Journal Contents Listing