Towards a Peer-to-Peer Bandwidth Marketplace

  • Mihai Capotă
  • Johan Pouwelse
  • Dick Epema
Conference paper

DOI: 10.1007/978-3-642-45249-9_20

Part of the Lecture Notes in Computer Science book series (LNCS, volume 8314)
Cite this paper as:
Capotă M., Pouwelse J., Epema D. (2014) Towards a Peer-to-Peer Bandwidth Marketplace. In: Chatterjee M., Cao J., Kothapalli K., Rajsbaum S. (eds) Distributed Computing and Networking. ICDCN 2014. Lecture Notes in Computer Science, vol 8314. Springer, Berlin, Heidelberg

Abstract

Peer-to-peer systems are a popular means of transferring files over the Internet, accounting for a third of the upload bandwidth of end users as of 2013. However, recent studies have highlighted that peer-to-peer systems are affected by a lack of balance between the supply and demand of bandwidth. This imbalance stems from the skewed popularity distribution of the files transfered in the system; newly released files may exhibit an undersupply of bandwidth while older ones may exhibit oversupply. In this work, we introduce a bandwidth marketplace for peers, with the aim of aligning supply and demand without the need for human intervention. Peers constantly monitor their performance and gossip with each other about undersupplied files. Peers with idle upload bandwidth that learn about an undersupplied file can autonomously start a special help mode download, with the goal of supplying as much upload bandwidth as possible to the other peers. We present an analytical model of help mode downloading and derive from it bounds for the performance of helper peers. Furthermore, we evaluate a recent existing implementation of help mode in Libtorrent, a popular BitTorrent library. Our tests show that Libtorrent help mode is effective at alleviating undersupply, although its performance relative to our model can be improved.

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

Copyright information

© Springer-Verlag Berlin Heidelberg 2014

Authors and Affiliations

  • Mihai Capotă
    • 1
  • Johan Pouwelse
    • 1
  • Dick Epema
    • 1
  1. 1.Delft University of TechnologyThe Netherlands

Personalised recommendations