Criticality-based Analysis and Design of Unstructured Peer-to-Peer Networks as "Complex Systems"

  • Authors:
  • Farnoush Banaei-Kashani;Cyrus Shahabi

  • Affiliations:
  • -;-

  • Venue:
  • CCGRID '03 Proceedings of the 3st International Symposium on Cluster Computing and the Grid
  • Year:
  • 2003

Quantified Score

Hi-index 0.00

Visualization

Abstract

Due to enormous complexity of the unstructured peer-to-peernetworks as large-scale, self-configure, and dynamicsystems, the models used to characterize these systems areeither inaccurate, because of oversimplification, or analyticallyinapplicable, due to their high complexity. By recognizingunstructured peer-to-peer networks as "complex systems",we employ statistical models used before to characterizecomplex systems for formal analysis and efficient designof peer-to-peer networks. We provide two examples ofapplication of this modeling approach that demonstrate itspower. For instance, using this approach we have been ableto formalize the main problem with normal flooding search,propose a remedial approach with our probabilistic floodingtechnique, and find the optimal operating point for probabilisticflooding rigorously, such that it improves scalabilityof the normal flooding by 99%.