Competitive On-Line Algorithms for Distributed Data Management

  • Authors:
  • Carsten Lund;Nick Reingold;Jeffery Westbrook;Dicky Yan

  • Affiliations:
  • -;-;-;-

  • Venue:
  • SIAM Journal on Computing
  • Year:
  • 1999

Quantified Score

Hi-index 0.01

Visualization

Abstract

Competitive on-line algorithms for data management in a network of processors are studied in this paper. A data object such as a file or a page of virtual memory is to be read and updated by various processors in the network. The goal is to minimize the communication costs incurred in serving a sequence of such requests. Distributed data management on important classes of networks---trees and bus-based networks---are studied. Optimal algorithms with constant competitive ratios and matching lower bounds are obtained. Our algorithms use different interesting techniques, such as work functions [Chrobak and Larmore, Proc. DIMACS Workshop on On-Line Algorithms, AMS, 1991, pp. 11--64] and "factoring."