Content distribution in heterogenous video-on-demand p2p networks with ARIMA forecasts

  • Authors:
  • Chris Loeser;Gunnar Schomaker;André Brinkmann;Mario Vodisek;Michael Heidebuer

  • Affiliations:
  • Institute of Computer Science, University of Paderborn, Germany;Institute of Computer Science, University of Paderborn, Germany;Institute of Computer Science, University of Paderborn, Germany;Institute of Computer Science, University of Paderborn, Germany;Institute of Computer Science, University of Paderborn, Germany

  • Venue:
  • ICN'05 Proceedings of the 4th international conference on Networking - Volume Part II
  • Year:
  • 2005

Quantified Score

Hi-index 0.00

Visualization

Abstract

Peer to peer applications have gained high popularity in the past years. In particular, P2P media streaming architectures have attracted much attention, so that audio and video sharing is causing a large fraction of the Internet traffic today. In this paper we introduce a new peer to peer architecture that focuses on distributed video on demand file sharing and that is based on point-to-point file delivery between the peers. The P2P architecture includes dynamic data distribution and replication schemes that are able to guarantee a fair load balancing among the peers. This load balancing enables the P2P architecture to avoid hot spots inside the distribution network and to ensure a nearly optimal throughput. A main component of the P2P architecture is an ARIMA based forecasting module, that is able to predict the access probability of individual files. This forecasting module has an impact on the location of documents concerning the characteristic of peers and in addition is used to control the number of replicas of each file. In this paper we present some simulation results indicating not only the feasibility of this architectural approach but also the benefits resulting from dynamic content distribution.