A New Algorithm for the Decentralized Aggregation Problem

Authors:
Akhil Kumar
Affiliations:
College of Business, CB 419, University of Colorado, Boulder, CO 80309
Venue:
Distributed and Parallel Databases
Year:
1999

Citing 7
Cited 0

Efficient decentralized consensus protocols

IEEE Transactions on Software Engineering
Semantics based transaction management techniques for replicated data

SIGMOD '88 Proceedings of the 1988 ACM SIGMOD international conference on Management of data
A N algorithm for mutual exclusion in decentralized systems

ACM Transactions on Computer Systems (TOCS)
An O(nlog n) Unidirectional Algorithm for the Circular Extrema Problem

ACM Transactions on Programming Languages and Systems (TOPLAS)
Nonblocking commit protocols

SIGMOD '81 Proceedings of the 1981 ACM SIGMOD international conference on Management of data
Distributed Computing on Regular Networks with Anonymous Nodes

IEEE Transactions on Computers
The Demarcation Protocol: A Technique for Maintaining Linear Arithmetic Constraints in Distributed Database Systems

EDBT '92 Proceedings of the 3rd International Conference on Extending Database Technology: Advances in Database Technology

Quantified Score

Hi-index	0.00

Visualization

Abstract

This paper describes a new algorithm for solving thedistributed aggregation problem in which individual valuesfrom N nodes have to be aggregated in an associative andcommutative manner, and the final result must be either computedat all sites or communicated to them. Our algorithm has a message-delay product ofΘ(dN(N^)) where N is the number of nodes, and d is a parameterthat corresponds to the number of dimensions of the hypercubeinto which the nodes are logically organized.This reflects an improvement upon existing algorithms by a factor of d. The algorithm can be initiated by any node, and works by creatingmultiple parallel streams that collect partial aggregates andfinally converge at a node to compute the final result.