SIParCS 2020 - Thomas Johnson III

CESM Load Balancing Development: Python Scripts for Workflow

Recorded Talk

Thanks to the enormous quantities of data that have been acquired concerning the climate systems of the Earth, software has been developed to model simulations of said climate systems.  One such software package is the Community Earth System Model, or CESM. CESM has been constructed and refined over the years by experts in the earth sciences and/or computer science field to better serve its users. As further research is carried out using CESM, complementary software has been made available to amplify the capabilities of CESM. This includes adding more components that can simulate different aspects of the Earth’s climate system. One such addition is a load balancer that accepts timing related data to generate from previously run CESM models to provide optimal parameters to build and run CESM models to prevent waste of resources. Prevention of wasted resources saves users computing resources as well as providing measures to allocate more saved computing resources for NCAR to more research endeavors. Said timing data has to be added manually at present. CESM can waste computing resources from allocating too many or too few resources to efficiently build and run CESM models. A new workflow for the CESM and load balancing process has been developed to allow for building and running CESM models and then executing the CESM load balancer through one process. Further modifications can be made to increase configuration options for CESM which opens opportunities for further flexibility for users when building and running CESM models.

SIParCS Project Partner: Soudeh Kamali

Mentors: Sheri Mickelson, Brian Dobbins, John Dennis

Slides