Load-driven neighbourhood reconfiguration of Gnutella overlay

  • Authors:
  • Evangelos Pournaras;Georgios Exarchakos;Nick Antonopoulos

  • Affiliations:
  • Vrije Universiteit Amsterdam, Department of Computer Science, De Boelelaan 1081a, 1081 HV, Amsterdam, The Netherlands;Computing Department, School of Electronics and Physical Sciences, University of Surrey, Guildford, Surrey GU2 7XH, United Kingdom;Computing Department, School of Electronics and Physical Sciences, University of Surrey, Guildford, Surrey GU2 7XH, United Kingdom

  • Venue:
  • Computer Communications
  • Year:
  • 2008

Quantified Score

Hi-index 0.24

Visualization

Abstract

Unstructured P2P networks support distributed applications whose workload may vary significantly over time and between nodes. Self-optimizing systems try to keep the load in the network balanced despite the frequent load fluctuations. Several P2P systems exhibit a number of related features but fail to avoid centralisation under high-load situations. ERGO aims to balance the overloaded nodes by rewiring some of their incoming links to underloaded ones via a set of interconnected servers which index the underloaded nodes. In two simulated environments, ERGO load-balancing on Gnutella network increases the balanced nodes and network availability by preserving its efficiency and even reducing its messages.