Trading bit, message, and time complexity of distributed algorithms

  • Authors:
  • Johannes Schneider;Roger Wattenhofer

  • Affiliations:
  • Computer Engineering and Networks Laboratory, ETH Zurich, Zurich, Switzerland;Computer Engineering and Networks Laboratory, ETH Zurich, Zurich, Switzerland

  • Venue:
  • DISC'11 Proceedings of the 25th international conference on Distributed computing
  • Year:
  • 2011

Quantified Score

Hi-index 0.00

Visualization

Abstract

We present tradeoffs between time complexity t, bit complexity b, and message complexity m. Two communication parties can exchange Θ(mlog(tb/m2)+b) bits of information for m bt and Θ(b) for m ≥ √bt. This allows to derive lower bounds on the time complexity for distributed algorithms as we demonstrate for the MIS and the coloring problems. We reduce the bit-complexity of the state-of-the art O(Δ) coloring algorithm without changing its time and message complexity. We also give techniques for several problems that require a time increase of tc (for an arbitrary constant c) to cut both bit and message complexity by Ω(log t). This improves on the traditional time-coding technique which does not allow to cut message complexity.