Efficient execution of scientific computation on geographically distributed clusters

Authors:
Eduardo Argollo;Dolores Rexachs;Fernando G. Tinetti;Emilio Luque
Affiliations:
Computer Science Department, Universitat Autònoma de Barcelona, Spain;Computer Science Department, Universitat Autònoma de Barcelona, Spain;Fac. de Informática, Inv. Asistente CICPBA, Universidad Nacional de La Plata, Argentina;Computer Science Department, Universitat Autònoma de Barcelona, Spain
Venue:
PARA'04 Proceedings of the 7th international conference on Applied Parallel Computing: state of the Art in Scientific Computing
Year:
2004

Citing 7
Cited 1

A set of level 3 basic linear algebra subprograms

ACM Transactions on Mathematical Software (TOMS)
The cache performance and optimizations of blocked algorithms

ASPLOS IV Proceedings of the fourth international conference on Architectural support for programming languages and operating systems
Libraries for linear algebra

High performance computing
A high-performance, portable implementation of the MPI message passing interface standard

Parallel Computing
Matrix Multiplication on Heterogeneous Platforms

IEEE Transactions on Parallel and Distributed Systems
Architectures for an Efficient Application Execution in a Collection of HNOWS

Proceedings of the 9th European PVM/MPI Users' Group Meeting on Recent Advances in Parallel Virtual Machine and Message Passing Interface
The master-slave paradigm with heterogeneous processors

IEEE Transactions on Parallel and Distributed Systems

Tuning application in a multi-cluster environment

Euro-Par'06 Proceedings of the 12th international conference on Parallel Processing

Quantified Score

Hi-index	0.00

Visualization

Abstract

To achieve data intensive computation, the joining of geographically distributed heterogeneous clusters of workstations through the Internet can be an inexpensive approach. To obtain effective collaboration in such a collection of clusters, overcoming processors and networks heterogeneity, a system architecture was defined. This architecture and a model able to predict application performance and to help its design is described. The matrix multiplication algorithm is used as a benchmark and experiments are conducted over two geographically distributed heterogeneous clusters, one in Brazil and the other in Spain. The model obtained over 90% prediction accuracy in the experiments.