Authors: Maeva Antoine; Fabrice Huet
Addresses: University of Nice Sophia Antipolis, CNRS I3S UMR 7271, 2000 route des Lucioles, Batiment Euclide A, 06900 Sophia Antipolis, France ' University of Nice Sophia Antipolis, CNRS I3S UMR 7271, 2000 route des Lucioles, Batiment Euclide A, 06900 Sophia Antipolis, France
Abstract: Hash functions are at the heart of data insertion and retrieval in DHT-based overlays. However, a standard hash function destroys the natural ordering of data. To perform efficient range queries processing, more and more systems opt for an order-preserving hash function to place data. Unlike a standard hash function, this technique cannot evenly distribute data among peers. We propose a novel approach to improve the data dissemination using several order-preserving hash functions. We describe a protocol that allows an overloaded peer to change its hash function at runtime. Then, we show that all peers should not necessarily use the same hash function in an overlay to insert or look for an item. Finally, we demonstrate through simulations that this strategy greatly improves the dissemination of data items. To our knowledge, this is the first load balancing solution using multiple order-preserving hash functions in structured overlays.
Keywords: hash functions; load balancing; structured overlays; skewed data; lexicographic order; peer-to-peer; P2P; semantic web; resource description framework; RDF.
International Journal of Communication Networks and Distributed Systems, 2018 Vol.20 No.1, pp.36 - 59
Available online: 10 Nov 2017 *Full-text access for editors Access for subscribers Free access Comment on this article