Networking

FRANTIC: A Fast Reference-Based Algorithm for Network Tomography Via Compressive Sensing

Free registration required

Executive Summary

The authors study the problem of link and node delay estimation in undirected networks when at most k out of n links or nodes in the network is congested. Their approach relies on end-to-end measurements of path delays across pre-specified paths in the network. They present a class of algorithms that they call FRANTIC. The FRANTIC algorithms are motivated by compressive sensing; however, unlike traditional compressive sensing, the measurement design here is constrained by the network topology and the matrix entries are constrained to be positive integers.

  • Format: PDF
  • Size: 1054.72 KB