Network patterns in cfengine and scalable data aggregation

  • Authors:
  • Mark Burgess;Matthew Disney;Rolf Stadler

  • Affiliations:
  • Oslo University College;Oslo University College;KTH Royal Institute of Technology, Stockholm

  • Venue:
  • LISA'07 Proceedings of the 21st conference on Large Installation System Administration Conference
  • Year:
  • 2007

Quantified Score

Hi-index 0.00

Visualization

Abstract

Network patterns are based on generic algorithms that execute on tree-based overlays. A set of such patterns has been developed at KTH to support distributed monitoring in networks with non-trivial topologies. We consider the use of this approach in logical peer networks in cfengine as a way of scaling aggregation of data to large organizations. Use of 'deep' network structures can lead to temporal anomalies. We show how to minimize temporal fragmentation during data aggregation by using time offsets and what effect these choices might have on power consumption. We offer proof of concept for this technology to initiate either multicast or inverse multicast pulses through sensor networks.