Distributed random digraph transformations for peer-to-peer networks

  • Authors:
  • Peter Mahlmann;Christian Schindelhauer

  • Affiliations:
  • University of Paderborn, Germany;University of Freiburg, Germany

  • Venue:
  • Proceedings of the eighteenth annual ACM symposium on Parallelism in algorithms and architectures
  • Year:
  • 2006

Quantified Score

Hi-index 0.00

Visualization

Abstract

We present a local random graph transformation for weakly connected multi-digraphs with regular out-degree which produces every such graph with equal probability. This operation, called Pointer-Push&Pull, changes only two neighboring edges. Such an operation is highly desirable for a peerto-peer network to establish and maintain well connected expander graphs as reliable and robust network backbone. The Pointer-Push&Pull operation can be used in parallel without central coordination and each operation involves only two peers which have to exchange two messages, each carrying the information of one edge only.We show that a series of random Pointer-Push&Pull operations eventually leads to a uniform probability distribution over all weakly connected out-regular multi-digraphs. Depending on the probabilities used in the operation this uniform probability distribution either refers to the set of all weakly connected out-regular multi-digraphs or to the set of all weakly connected out-regular edge-labeled multidigraphs. In multi-digraphs multiple edges or self-loops may occur. In an out-regular digraph each node has the same number of outgoing edges.For this, we investigate the Markov-Process defined by the Pointer-Push&Pull operation over the set of all weakly connected multi-digraphs. We show that a Pointer-Push&Pull operation -- although preserving weak connectivity only -- can reach every weakly connected multi-digraph. The main argument follows from the symmetry of the Markov-Process described by the Pointer-Push&Pull operation over the set of all weakly connected out-regular multi-digraphs.