Enhancing Load Balancing in a Data-Parallel GSM Network Simulation through Application-Specific Information

Authors:
Pentti Huttunen;Jouni Ikonen;Jari Porras
Affiliations:
-;-;-
Venue:
PARA '02 Proceedings of the 6th International Conference on Applied Parallel Computing Advanced Scientific Computing
Year:
2002

Citing 8
Cited 0

Models of machines and computation for mapping in multicomputers

ACM Computing Surveys (CSUR)
Competitive analysis of network load balancing

Journal of Parallel and Distributed Computing
Theoretical analysis of the heterogeneous dynamic load-balancing problem using a hydrodynamic approach

Journal of Parallel and Distributed Computing
Customized dynamic load balancing for a network of workstations

Journal of Parallel and Distributed Computing
Analysis of load average and its relationship to program run time on networks of workstations

Journal of Parallel and Distributed Computing
A load index for dynamic load balancing

ACM '86 Proceedings of 1986 ACM Fall joint computer conference
Load balancing strategies for distributed memory machines

Parallel and distributed processing for computational mechanics
Adaptive Multivariate Integration using MPI

HIPC '97 Proceedings of the Fourth International Conference on High-Performance Computing

Quantified Score

Hi-index	0.00

Visualization

Abstract

Load balancing plays an important role in achieving good performance in any parallel application. In order to produce a proper load distribution and to minimize overhead, a load balancing algorithm needs to take into account characteristics of the application. Three load balancing algorithms that have been enhanced to optimize the performance of a data-parallel simulator are presented. The enhanced load balancing algorithms are provided with additional information about the workunits, such as an estimate of computational requirements. Results show that the additional information has a significant impact on the load balancing algorithms. The algorithms are capable of producing a more optimal load balance, which leads to the improved performance of an application. A closer study of the overhead created by the processing of the information by the load balancing algorithms indicates that the improvement in the performance significantly outweighs the negative impact of the overhead in most cases.