This new funding will be given to MCS over a three year period as part of a series of research grants for data management and visualization. The awards will fund four major projects.
The first involves research into how to accurately and reliably visualize complex data consisting of multiple nonuniform domains and/or data types. In collaboration with three universities—Utah, Notre Dame, and Nebraska–Lincoln—the researchers will develop implicit continuous representations for visualization of disparate data.
“Having such a uniform data representation will allow us to compare data sources in order to explain scientific conclusions,” said Tom Peterka, a computer scientist in the MCS division and principal investigator of the project.
The second project, a collaboration with the National Renewable Energy Laboratory (lead institution) and Ohio State University, seeks to design visualization models to help users understand the outcome of ensemble simulations. Currently researchers are limited by the lack of tools that integrate parameter and uncertainty investigations with effective approaches for decision-making. Franck Cappello, a senior computer scientist in the MCS division, will lead the investigation of variational models.
“Our goal is to refine our understanding of a simulation’s sensitivity to certain input parameters and use that information for performance tuning,” Cappello said.
The other two newly funded projects involve next-generation data management. One, led by Robert Ross, a senior computer scientist in Argonne’s MCS division, seeks to develop extensions to the Mochi data navigation tool. An important aspect of Mochi is composition: common capabilities such as communication, data storage, and concurrency management.
“Our aim is to extend this approach, enabling integration of smart devices in data services and facilitating service elasticity for future high-performance computing platforms and workloads,” Ross said. The work will be done in collaboration with Los Alamos National Laboratory, Carnegie Mellon University and the New Mexico Consortium.
The second data management project is a collaboration with Brookhaven National Laboratory (lead institution) and Texas State University. Its aim is to develop novel approaches for capturing, fusing, storing and organizing the rich multimodal information – both data and metadata – in order to enable reproducibility of hybrid workflows at scale.
“Using the FAIR (Findable, Accessible, Interoperable, and Reusable) guiding principles for scientific data management makes this information easier to organize and query, which enables us to explore efficient reproducibility techniques to analyze divergence between repeated runs with respect to both results and performance,” said Bogdan Nicolae, a computer scientist in Argonne’s MCS division and co-investigator of the newly funded project.