Partial read from peer-to-peer databases

  • Authors:
  • Farnoush Banaei-Kashani;Cyrus Shahabi

  • Affiliations:
  • University of Southern California, Department of Computer Science, Information Laboratory (InfoLab), Los Angeles, CA 90089-0781, USA;University of Southern California, Department of Computer Science, Information Laboratory (InfoLab), Los Angeles, CA 90089-0781, USA

  • Venue:
  • Computer Communications
  • Year:
  • 2008

Quantified Score

Hi-index 0.24

Visualization

Abstract

In this paper, we propose Scoop, a mechanism to implement the ''partial read operation'' for peer-to-peer databases. A peer-to-peer database is a database that its relations are horizontally fragmented and distributed among the nodes of a peer-to-peer network. The partial read operation is a data retrieval operation required for approximate query processing in peer-to-peer databases. A partial read operation answers to @b-queries: given @b@?[0,1] and a relation R, a fraction @b of the tuples in R must be retrieved from the database to answer a @b-query. Despite the simplicity of the @b-query, due to the distributed, evolving and autonomous nature of the peer-to-peer databases correct and efficient implementation of the partial read operation is challenging. Scoop is designed based on an epidemic dissemination algorithm. We model the epidemic dissemination as a percolation problem and by rigorous percolation analysis tune Scoop per-query and on-the-fly to answer @b-queries correctly and efficiently. We prove the correctness of Scoop by theoretical analysis, and verify the efficiency of Scoop in terms of query cost and query time via extensive simulation.