Robust topology control protocols

  • Authors:
  • Sukumar Ghosh;Kevin Lillis;Saurav Pandit;Sriram Pemmaraju

  • Affiliations:
  • The University of Iowa, Iowa City, IA;The University of Iowa, Iowa City, IA;The University of Iowa, Iowa City, IA;The University of Iowa, Iowa City, IA

  • Venue:
  • OPODIS'04 Proceedings of the 8th international conference on Principles of Distributed Systems
  • Year:
  • 2004

Quantified Score

Hi-index 0.00

Visualization

Abstract

Topology control protocols attempt to reduce the energy consumption of nodes in an ad-hoc wireless network while maintaining sufficient network connectivity. Topology control protocols with various features have been proposed, but they all lack robustness and are extremely sensitive to faulty information from neighbors. For example, the XTC protocol (R. Wattenhofer and A. Zollinger, XTC: A practical topology control algorithm for ad-hoc networks, WMAN 2004) can be forced to construct a disconnected network even if two nodes in the network receive slightly faulty distance information from one neighbor each. A key step in most localized topology control protocols is one in which each node establishes a total ordering on its set of neighbors based on information received from them. In this paper, we propose a metric for robustness of localized topology control protocols and define an r-robust topology control protocol as one that returns a correct output network even when its neighborhood orderings have been modified by up to r–1 adjacent swaps by a malicious adversary. We then modify XTC in a simple manner to derive a family of r-robust protocols for any r 1. The price we pay for increased robustness is in terms of decreased network sparsity; however we can bound this decrease and we show that in transforming XTC from a 1-robust protocol (which it trivially is) into an r-robust protocol, the maximum vertex degree of the output network increases by a factor of $O(\sqrt{r})$. An extremely pleasant side-effect of our design is that the output network is both $\Omega(\sqrt{r})$-edge connected and $\Omega(\sqrt{r})$-vertex connected provided the input network is. Thus ensuring robustness of the protocol seems to give fault-tolerance of the output for free. Our r-robust version of XTC is almost as simple and practical as XTC and like XTC it only involves 2 rounds of communication between a node and its neighbors.