VECPAR'06 - Seventh International Meeting on High Performance Computing for Computational Science
vecpar.fe.up.pt/2006 | vecpar2006@fe.up.pt
Grid Computing for Mesoscale Climatology: Experimental Comparison of Three Platforms
Haroldo F. de Campos Velho (LAC/INPE)
A. J. Preto (LAC/INPE)
S. Stephany (LAC/INPE)
E. R. Rodrigues (LAC/INPE)
J. Panetta (CPTEC/INPE)
E. S. Almeida (CPTEC/INPE)
R. P. Souto (CPTEC/INPE)
P. O. A. Navaux (II/UFRGS)
T. A. Diverios (II/UFRGS)
N. Maillard (II/UFRGS)
P. L. Silva Dias (IAG/USP)
Abstract:
A good number of Grid middlewares have been developed recently, yet their effective use for real world applications is seldom validated. Meanwhile, high-performance computing codes keep on requiring more resources than even large clusters can deliver. In this paper, the Gbrams project is presented, which aims at executing climatology simulations on a three tiers grid. Three different middlewares (Globus, OAR/CIGRI and OurGrid) are compared and some preliminary measurements are presented. The outcome will be the definition of a Computational Grid, adapted to run a mesoscale meteorological model (BRAMS) used in weather and climatology forecasting that may demand several hundreds of independent executions, one for each different initial condition. The proposed grid is employed to perform these executions for a given middleware. Each execution is scheduled as an MPI job in a particular cluster that is part of the grid.
Keywords:
User experience in deploying grids and testbeds , Grid and cluster management
 
Logos Universidade Federal do Rio de Janeiro - Coordenação dos Programas de Pós-graduação de Engenharia Instituto Nacional de Matemática Pura e Aplicada Rio de Janeiro | Brazil | 2006 | July | 10 11 12 13