Optimising multicast structures for grid computing

  • Authors:
  • Gill Waters;John Crawford;Sei Guan Lim

  • Affiliations:
  • Computing Laboratory, University of Kent, Canterbury, Kent CT2 7NF, UK;Computing Laboratory, University of Kent, Canterbury, Kent CT2 7NF, UK;Computing Laboratory, University of Kent, Canterbury, Kent CT2 7NF, UK

  • Venue:
  • Computer Communications
  • Year:
  • 2004

Quantified Score

Hi-index 0.24

Visualization

Abstract

This paper introduces a flexible new method of constructing hierarchical multicast structures suitable for supporting large-scale GRID applications. Hierarchical multicast trees are constructed by repeated application of clustering algorithms that partition the members of a large application community to form a layered hierarchy of clusters of users. The hierarchies are examples of application overlay networks that do not rely on network layer facilities for multicast transmission. The centralized method described applies clustering based either on the geographical location of users or by using more standard network topology measures. Our results show favourable performance when compared with the tree-building algorithm of the NICE protocol for overlay networks. Both NICE and our clustering approach build overlay networks which offer a compromise between the 'stress' on network links and the 'stretch' or increase in delay to users caused by the application layer overlay.