Comprehensive climate system modeling on massively parallel computers
A better understanding of both natural and human induced changes to the Earth`s climate is necessary for policy makers to make informed decisions regarding energy usage and other greenhouse gas producing activities. To achieve this, substantial increases in the sophistication of climate models are required. Coupling between the climate subsystems of the atmosphere, oceans, cryosphere and biosphere is only now beginning to be explored in global models. the enormous computational expenses of such models is one significant factor limiting progress. A comprehensive climate system model targeted to distributed memory massively parallel processing (MPP) computers is under development at Lawrence Livermore National Laboratory. This class of computers promises the computational power to permit the timely execution of climate models of substantially more sophistication than current generation models. Our strategy for achieving high performance on large numbers of processors is to exploit the multiple layers of parallelism naturally contained within highly coupled global climate models. The centerpiece of this strategy is the concurrent execution of multiple independently parallelized components of the climate system model. This methodology allows the assignment of an arbitrary number of processors to each of the major climate subsystems. Hence, a higher total number of processors may be efficiently used. Furthermore, load imbalances arising from the coupling of submodels may be minimized by adjusting the distribution of processors among the submodels.
- Research Organization:
- Lawrence Livermore National Lab. (LLNL), Livermore, CA (United States)
- Sponsoring Organization:
- USDOE, Washington, DC (United States)
- DOE Contract Number:
- W-7405-ENG-48
- OSTI ID:
- 492089
- Report Number(s):
- UCRL-JC-125558; CONF-970157-3; ON: DE97052677
- Resource Relation:
- Conference: 1997 Society of Computer Simulation (SCS) western multiconference, Phoenix, AZ (United States), 12-15 Jan 1997; Other Information: PBD: Oct 1996
- Country of Publication:
- United States
- Language:
- English
Similar Records
Climate systems modeling on massively parallel processing computers at Lawrence Livermore National Laboratory
A Framework to Analyze the Performance of Load Balancing Schemes for Ensembles of Stochastic Simulations