Adaptive Algorithms for Join Processing in Distributed Database Systems

  • Authors:
  • Peter Scheuermann;Eugene Inseok Chong

  • Affiliations:
  • Department of Electrical and Computer Engineering, Northwestern University, Evanston, IL 60208. E-mail: peters@ece.nwu.edu;New England R&D Center, Oracle Corporation, 1 Oracle Drive, Nashua, NH 03062. E-mail: echong@us.oracle.com

  • Venue:
  • Distributed and Parallel Databases
  • Year:
  • 1997

Quantified Score

Hi-index 0.00

Visualization

Abstract

Distributed query processing algorithms usually performdata reduction by using a semijoin program, but the problem withthese approaches is that they still require an explicit join of thereduced relations in the final phase. We introduce an efficientalgorithm for join processing in distributed database systems thatmakes use of bipartite graphs in order to reduce data communicationcosts and local processing costs. The bipartite graphs represent thetuples that can be joined in two relations taking also into accountthe reduction state of the relations. This algorithm fully reducesthe relations at each site. We then present an adaptive algorithm forresponse time optimization that takes into account the systemconfiguration, i.e., the additional resources available and the datacharacteristics, in order to select the best strategy for responsetime minimization. We also report on the results of a set ofexperiments which show that our algorithms outperform a number of therecently proposed methods for total processing time and response timeminimization.