Distributed PageRank computation with link failures

Abstract

The Google search engine employs the so-called PageRank algorithm to rank the search results by quantifying the importance of each web page. In this paper, we continue our recent work on distributed randomized computation of PageRank, where the pages locally determine their values by communicating with linked pages. In particular, we propose a distributed randomized algorithm with limited information, where only part of the linked pages is required to be contacted. This is useful to enhance flexibility and robustness in computation and communication. © 2010 Springer-Verlag Berlin Heidelberg.