Home | Poster Submission | Posters | Archives | Contacts |
A Workflow Management System for Automating Weather and Climate Simulations Christopher W. Harrop, Ligia Bernardet, Mark Govett, Jeff S. Smith, Stephen Weygandt One of the challenges of modeling and simulating weather and climate phenomena is the laborious process of running the various codes that are required to carry out a simulation. For example, to perform a single end-to-end run of a weather simulation, a scientist must often run several data preprocessing and assimilation programs, the model code, several postprocessing programs, codes to produce visualizations, and at least one forecast verification program. These codes often have complex interdependencies and if the simulation is run in real time, such as in an operational setting, the various programs must run at specific times of the day and must finish before the resulting forecasts become irrelevant. Another complication is that many of these codes must run on High Performance Computing (HPC) platforms that are not always reliable and are notoriously difficult for scientists to use. The onerous task of properly managing the execution of all the various simulation codes is exacerbated even further by the fact that hundreds to thousands of end-to-end runs of the simulation are required to complete many experiments. The net effect of all these obstacles is that scientists often become overwhelmed by the process of running their simulations and have little time left to pursue the scientific questions they set out to investigate. In collaboration with the Development Testbed Center (DTC), we have developed a scientific Workflow Management System (WFMS) to address these above challenges. |