Keywords

Accelerated Climate Modeling for Energy (ACME), ACME Land Model (ALM); In situ data visualization; common communication interface (CCI); High performance computing (HPC).

Location

Session D3: Advances in Environmental - Decision Support - Software Systems

Start Date

12-7-2016 10:50 AM

End Date

12-7-2016 11:10 AM

Abstract

With the ever increasing precision and complexity of various environmental models, writing large volumes of datpeiy@ornl.gova to disk for post-processing becomes a huge bottleneck, and simulation modules diagnostic and tracking become difficult for scientists as well. In this paper, we propose a loosely coupled in situ visualization framework to address these problems when working with a representative land model within the Accelerated Climate Modeling for Energy (ACME). ACME land model (ALM) simulates energy, water, carbon nitrogen fluxes and state variables for various land surfaces. Loosely coupled in situ framework means simulation and visualization communicate through network, while tightly coupled means they are integrated and run on the same nodes. Built with a high performance common communication interface (CCI) and a customized data visualization environment (DVE), we demonstrated our lightweight framework for a single site ALM simulation case running on Titan supercomputer. Our framework makes real-time simulation monitoring possible, and the customized DVE provides scientists an intuitive and useful way to explore results. We believe our flexible and performant in situ framework can be readily adapted to a wide range of simulations, expediting scientific discoveries. For future work, we will expand our current framework to include task scheduling, enabling multitask visualization of large-scale parallel environmental model simulations.

Share

COinS
 
Jul 12th, 10:50 AM Jul 12th, 11:10 AM

In situ Data Visualization of ACME Land Model Simulations

Session D3: Advances in Environmental - Decision Support - Software Systems

With the ever increasing precision and complexity of various environmental models, writing large volumes of datpeiy@ornl.gova to disk for post-processing becomes a huge bottleneck, and simulation modules diagnostic and tracking become difficult for scientists as well. In this paper, we propose a loosely coupled in situ visualization framework to address these problems when working with a representative land model within the Accelerated Climate Modeling for Energy (ACME). ACME land model (ALM) simulates energy, water, carbon nitrogen fluxes and state variables for various land surfaces. Loosely coupled in situ framework means simulation and visualization communicate through network, while tightly coupled means they are integrated and run on the same nodes. Built with a high performance common communication interface (CCI) and a customized data visualization environment (DVE), we demonstrated our lightweight framework for a single site ALM simulation case running on Titan supercomputer. Our framework makes real-time simulation monitoring possible, and the customized DVE provides scientists an intuitive and useful way to explore results. We believe our flexible and performant in situ framework can be readily adapted to a wide range of simulations, expediting scientific discoveries. For future work, we will expand our current framework to include task scheduling, enabling multitask visualization of large-scale parallel environmental model simulations.