Parallel hash join algorithms for dynamic load balancing in a shared disks cluster

Authors:
Aekyung Moon;Haengrae Cho
Affiliations:
Software Robot Research Team, ETRI, Taejon, Republic of Korea;Department of Computer Engineering, Yeungnam University, Gyungsan, Gyungbuk, Republic of Korea
Venue:
ICCSA'06 Proceedings of the 2006 international conference on Computational Science and Its Applications - Volume Part V
Year:
2006

Citing 8
Cited 0

A performance evaluation of four parallel join algorithms in a shared-nothing multiprocessor environment

SIGMOD '89 Proceedings of the 1989 ACM SIGMOD international conference on Management of data
Dynamic Data Reallocation for Skew Management inShared-Nothing Parallel Databases

Distributed and Parallel Databases
Database Systems Concepts

Database Systems Concepts
Query Processing in Parallel Relational Database Systems

Query Processing in Parallel Relational Database Systems
Dynamic and Load-balanced Task-Oriented Datbase Query Processing in Parallel Systems

EDBT '92 Proceedings of the 3rd International Conference on Extending Database Technology: Advances in Database Technology
Performance Comparison of Pipelined Hash Joins on Workstation Clusters

HiPC '02 Proceedings of the 9th International Conference on High Performance Computing
Cluster Based Hybrid Hash Join: Analysis and Evaluation

CLUSTER '02 Proceedings of the IEEE International Conference on Cluster Computing
Oracle Real Application Clusters

Oracle Real Application Clusters

Quantified Score

Hi-index	0.00

Visualization

Abstract

Most of previous parallel join algorithms assume a shared nothing (SN) cluster, where each database partition is owned by a single processing node. While SN cluster can interconnect a large number of nodes and support a geographically distributed environment, it may suffer from poor facility for load balancing and system availability compared to a shared disks sharing (SD) cluster. In this paper, we first propose a dynamic load balancing strategy by exploiting the characteristics of SD cluster. Then we parallelize conventional hash join algorithms using the dynamic load balancing strategy. We also explore the performance of parallel join algorithms using a simulation model of SD cluster. The experiment results show that the proposed parallel join algorithms can achieve higher potential for dynamic load balancing with the inherent flexibility of SD cluster.