Bounding communication cost in dynamic load balancing of distributed hash tables

  • Authors:
  • Marcin Bienkowski;Miroslaw Korzeniowski

  • Affiliations:
  • Institute of Computer Science, University of Wroclaw, Poland;International Graduate School of Dynamic Intelligent Systems, University of Paderborn, Germany

  • Venue:
  • OPODIS'05 Proceedings of the 9th international conference on Principles of Distributed Systems
  • Year:
  • 2005

Quantified Score

Hi-index 0.00

Visualization

Abstract

In Peer-to-Peer networks based on consistent hashing and ring topology each peer is responsible for an interval chosen (pseudo-) randomly on a circle. The topology of the network, the communication load and the amount of data a peer stores depends heavily on the length of its interval. Additionally, peers are allowed to join the network or to leave it at any time. Such operations can destroy the balance of the network even if all the intervals had equal lengths in the beginning. This paper deals with the task to keep such a system balanced, so that the lengths of intervals assigned to the peers differ at most by a constant factor. We propose a simple scheme which achieves this and its communication cost can be amortized against the cost of keeping the system connected. Our procedure requires O(logn) times more messages than any procedure maintaining the connectivity, even if the an oblivious adversary decides about the dynamics of the system. The scheme is a continuous process which does not have to be informed about the current size or possible imbalance in the network to start working. As a byproduct, we show how to compute a constant approximation of the current number of nodes n in the system, provided that we know an upper bound on logn.