Evaluating the impact of fanout and dimension on the performance of a hybrid model for multidimensional indexing in peer-to-peer m -ary tree network. (25th February 2015)

• Record Type: Journal Article
• Authors: Surati, Shivangi; Jinwala, Devesh C.; Garg, Sanjay
• Is Part Of: International journal of communication networks and distributed systems. Volume 14:Number 2(2015)
• Page Start: 185
• Page End: 201
• Publication Date: 2015-02-25
• Subjects: peer-to-peer networks -- distributed computing -- multidimensional indexing -- complex query search -- multidimensional data -- m -ary trees; Computer networks -- Periodicals; Telecommunication systems -- Periodicals; Electronic data processing -- Distributed processing -- Periodicals; 004.6
• Languages: English
• ISSNs: 1754-3916

Abstract

In the structured peer-to-peer (P2P) overlay networks, multidimensional indexing (MI) methods have emerged to address principally the efficiency concerns of complex queries. As per our observations, a hybrid model that combines P2P m -ary tree overlay network with traditional MI methods is preferable as it reduces the cost of search operations from log 2 N to logm N . We show an illustrative hybrid model in this paper that differentiates the nodes for storing routing information (routing nodes) and data objects (data nodes) independently. However, the search performance and the cost of update operations of this tree are dependent on two vital parameters viz. fanout of the tree ( m ) and dimensionality of the data object ( d ). In addition, variations of these parameters directly affect the space complexities of both the nodes. This paper discusses the experimentations and analyses: 1) appropriate tuning of these two parameters for efficient search performance and maintaining the update cost of the tree; 2) synchronisation of node sizes to overcome the memory issues related to two different types of nodes during node replacement.