A distributed selection algorithm and its expected communication complexity
Theoretical Computer Science
An efficient selection algorithm on the pyramid
Information Processing Letters
Efficient Distributed Selection with Bounded Messages
IEEE Transactions on Parallel and Distributed Systems
Java and Soap
An end-user perspective on file-sharing systems
Communications of the ACM
Tight Bounds for Shared Memory Systems Accessed by Byzantine Processes
DISC '02 Proceedings of the 16th International Conference on Distributed Computing
The Aleph Toolkit: Support for Scalable Distributed Shared Objects
CANPC '99 Proceedings of the Third International Workshop on Network-Based Parallel Computing: Communication, Architecture, and Applications
Parallel External Selection Algorithm on Distributed Memory Systems
ICA3PP '02 Proceedings of the Fifth International Conference on Algorithms and Architectures for Parallel Processing
An Adaptive Method for Unknown Distributions in Distributive Partitioned Sorting
IEEE Transactions on Computers
Hi-index | 0.00 |
This paper presents an efficient distributed multiple selection algorithm designed to select multiple keys simultaneously from different data sets which are distributed to many computers in a peer-to-peer system. The communication time is usually much longer than the computation time and is thus a major criterion for measuring the performance of a distributed algorithm. The objective of this algorithm is to reduce the number of communication messages. The algorithm makes use of statistical knowledge and results in a smaller communication overhead compared with existing algorithms. In this algorithm, each computer will select keys as pivots (candidates for the answers) according to statistical knowledge and transmit them to other computers in the system. Each computer will compare each pivot with key values in its local file and respond by transmitting ranks to the originating computer. The originating computer will calculate the global ranks and verify whether the pivots are the answers. Each computer will broadcast once sequentially in each round. This broadcasting process will be repeated until all answers are found. Complexity analyses are presented to provide theoretical upper bounds of this algorithm. The correctness of the theoretical predication is verified by experiments with 40 computers connected using Web technologies.