SIParCS 2020 - Soudeh Kamali
CESM Load Balancing Development-Optimizer Study
Understanding the global climate, how it has changed in the past, and how it is will change in the future, is a very important challenge that requires the systematic development of climate models. Modern climate models are incredibly complex applications which use coupled simulations. These coupled simulations contain standalone components such as the atmosphere, ocean, ice or land models that work together and communicate through a coupler to accurately simulate the entire earth system. Each of these components model different physical processes, and some calculations are more expensive than others. This makes balancing the work to be done across many processors difficult, and thus has typically been the domain of a few experts with first-hand knowledge about the systems and the models. A new approach is needed that does not rely on a limited number of individuals who are experts in the load balancing of these models. Instead we intend to develop an automatic load balancing capability. This project studies the potential automatic load balancing capability for the Community Earth System Model (CESM). This is done by performing short runs of each of the component models, and from their performance, calculating improved configuration.
SIParCS Project Partner: Thomas Johnson III
Mentors: Sheri Mickelson, Brian Dobbins, John Dennis