ChurnDetect: a gossip-based churn estimator for large-scale dynamic networks

  • Authors:
  • Andrei Pruteanu;Venkat Iyer;Stefan Dulman

  • Affiliations:
  • Embedded Software Group, Delft University of Technology, The Netherlands;Embedded Software Group, Delft University of Technology, The Netherlands;Embedded Software Group, Delft University of Technology, The Netherlands

  • Venue:
  • Euro-Par'11 Proceedings of the 17th international conference on Parallel processing - Volume Part II
  • Year:
  • 2011

Quantified Score

Hi-index 0.00

Visualization

Abstract

With the ever increasing scale of dynamic wireless networks (such as MANETs, WSNs, VANETs, etc.), there is a growing need for performing aggregate computations, such as online detection of network churn, via distributed, robust and scalable algorithms. In this paper we introduce the ChurnDetect algorithm, a novel solution to the distributed churn estimation problem. Our solution consists in a gossiping-based algorithm, which incorporates a periodic reset mechanism (introduced as DiffusionReset). The main difference with existing state-of-the-art is that ChurnDetect does not require nodes to advertise their departure from the network nor to detect neighbors leaving the network. In our solution, all the nodes are interacting with each other wirelessly, by using a gossip-alike approach, thus keeping the message complexity to a minimum. We only use easy accessible information (i.e., about new nodes joining the network) rather than presuming knowledge on nodes leaving the system since that is highly unfeasible for most distributed applications. We provide convergence proofs for ChurnDetect, and present a number of results based on simulations and implementation on our local testbed. We characterize the performance of the algorithm, showcasing its distributed light-weight characteristics. The analysis leads to the conclusion that ChurnDetect is an attractive alternative to existing work on online churn estimation for dynamic wireless networks.