Parallelizing a very high resolution climate model using clusters of workstations with PVM and performance and load balance analyses

Thumbnail Image
Supplemental Files
Date
1998
Authors
Wang, Hao
Prabhu, Gurpur
Takle, Eugene
Major Professor
Advisor
Committee Member
Journal Title
Journal ISSN
Volume Title
Publisher
Authors
Person
Research Projects
Organizational Units
Organizational Unit
Organizational Unit
Organizational Unit
Organizational Unit
Organizational Unit
Journal Issue
Is Version Of
Versions
Series
Department
Aerospace EngineeringAmes National LaboratoryComputer ScienceAgronomyGeological and Atmospheric Sciences
Abstract

Environment and climate change problems are very complicated, and their research and operational prediction heavily depend on powerful computer techniques. Even with today most powerful supercomputer, the climate and environmental models are still limited to very coarse resolution. In this paper, we report our recent effort in parallelizing our very-high-resolution numerical model systems. First, the mathematical equations, algorithms, and numerical schemes are designed and analyzed; then domain decomposition, data decomposition, and functional decomposition schemes are tested in our implementations on clusters of HP workstations and/or DEC Alpha stations with PVM; finally, the performance and load balance are analyzed. Shelterbelts cause significantly inhomogeneous computation distribution on the domain, therefore, common and easiest domain decomposition does not work well on our problem. Special care must be taken to treat computations around shelterbelts. With carefull design of algorithms, we found that cheap and still powerful workstations or PCs make it possible to run these models in clusters of workstations or PCs.

Comments

This proceeding was published as Wang, H., G. Prabhu, and E. S. Takle, 1998: "Parallelizing a very high resolution climate model using clusters of workstations with PVM and performance and load balance analyses." Proceedings of the International Conference on Parallel and Distributed Processing Techniques and Applications, CSREA Press. pp. 1762-1765. Posted with permission.

Description
Keywords
Citation
DOI
Source
Copyright
Thu Jan 01 00:00:00 UTC 1998