September 1, 2005
Journal Article

Load Balancing and Scalability of a Subgrid Orography Scheme in a Global Climate Model

Abstract

A subgrid orography scheme has been applied to the National Center for Atmospheric Research Community Atmosphere Model. The scheme applies all of the model column physics to each of up to eleven elevation classes within each grid cell. The distribution of the number of elevation classes in each grid cell is highly inhomogeneous. This could produce a serious load imbalance if the domain decomposition distributes grid cells evenly across processors. But since the distribution of classes is static, static load balancing can be used to distribute the elevation classes uniformly across processors. The load balancing is accomplished first by distributing the number of classes evenly within each node. The number of chunks on nodes is distributed uniformly across nodes and the dynamics-physics transpose cost is minimized by assigning chunks to nodes with the most dynamics grid cells from that chunk. Parallel efficiency with the subgrid scheme and load balancing exceeds parallel efficiency without the subgrid scheme for up to 128 processors. The load balancing across nodes decreases runtime by 10–30% depending on configuration.

Revised: October 27, 2005 | Published: September 1, 2005

Citation

Ghan S.J., and T.R. Shippert. 2005. Load Balancing and Scalability of a Subgrid Orography Scheme in a Global Climate Model. International Journal of High Performance Computing Applications 19, no. 3:237-245. PNNL-SA-43617.