P2P traffic localization by forcing packet loss

Hiep Hoang-Van, Koki Mizutani, Takumi Miyoshi, Olivier Fourmaux

Research output: Contribution to journalArticlepeer-review

1 Citation (Scopus)


Recently, peer-to-peer (P2P) traffic is increasing rapidly in volume day by day. One of the main causes is that most of P2P applications including file sharing and streaming applications often form overlay networks for exchanging data that are oblivious to the underlay network topology. As a result, they generate a large amount of inter-domain traffic causing higher cost for internet service providers (ISPs). This raises a problem of traffic localization. To optimize the cross-ISP/AS traffic, existing approaches focus on solving the problem on the application layer where each P2P application must be equipped with an additional protocol to obtain underlay network information from an "oracle" server or an additional locality-aware procedure to estimate location by itself. Therefore some modifications of application software are required for these approaches. In this paper, we propose a novel solution for addressing the problem, called PLS, forcing packet loss to each P2P packet based on geographical location of each destination at the network layer. Since PLS is implemented at a network router, no software modification is required. This proposal can be applied to all types of P2P applications in order to localize the traffic. The experiments evaluated on popular P2P streaming applications show that our proposed method significantly reduces the cross-domain traffic by suppressing the connections with faraway peers.

Original languageEnglish
Pages (from-to)251-259
Number of pages9
JournalInternational Journal of Networked and Distributed Computing
Issue number4
Publication statusPublished - 2013 Nov


  • P2P
  • P2PTV
  • Packet loss
  • Router aided
  • Traffic localization

ASJC Scopus subject areas

  • Computer Science Applications
  • Computer Networks and Communications


Dive into the research topics of 'P2P traffic localization by forcing packet loss'. Together they form a unique fingerprint.

Cite this