KITWARE INC — Department of Energy SBIR Phase II: 20c

Description

The promise of fusion energy is enormous - an energy system whose fuel is obtained fromsimple ingredients, and which yields zero carbon emissions to the atmosphere. However, realizing this vision requires undertaking one of the most challenging programs of scientific research and development that has ever been undertaken. Predictive simulation plays a crucial role in fusion energy research. Such simulations providea framework for the creation and evaluation of fusion experiments, and point the way towards viable designs. However, two key problems must be overcome if simulation is to remain effectiveas problem size and complexity increases. First, large-scale computing is producing large amounts of data and consequently data transfer is now the major bottleneck in most simulation workflows. Second, the complexity of multi-physics analyses requires the capability to easily configure software systems into workflows that can handle complex, multi-physics workflows. The scientific community has been working hard to develop the software infrastructure to tackle these challenges. In the proposed work, we will leverage and extend two software libraries that are key parts of this infrastructure: ADIOS and VTK-m. ADIOS is a system for fast, scalable, and portable IO and data transfer. It has been used to integrate diverse simulation and analysis toolsinto complex scientific workflows to address data scaling and multi-physics scientific challenges. However, one of the main difficulties in integrating such diverse tools is the incompatibility between various data models, with existing solutions being home-grown and ad-hoc. Our aim is to tacklethis challenge by integrating ADIOS with VTK-m. VTK-m is a high-performance system that supports a number of abstract data models. The flexibility of this data framework has already been demonstrated in various in situ analysis and visualization applications. Together the combined capabilities of ADIOS and VTK-m will enable scalable IO and support interoperable simulationtools that can be readily integrated into efficient simulation workflows. In our Phase I project, we have demonstrated the feasibility of this approach by developing a prototype system and applyingit to the XGC1 simulation data. This end-to-end prototype was limited to a single data model and a few algorithms. In our Phase II project, we propose to build a production ready end-to-end system that provides complex simulation and data analysis workflows to the fusion community. This system will support in situ, in transit as well as post hoc data analysis and visualization enabling scientists to leverage advanced analysis tools including VTK-m, VTK/ParaView and SciPy. While the work performed in this SBIR project is focused on fusion simulation, it is part of Kitware’s vision for advanced simulation.We have been building open source platforms which can be customized to meet specific customer needs. Such customization, as well as providing support, engaging in collaborative R&D, and performing technology integration, form the basis ofa service-driven business model. The tools developed here will enable Kitware to more easily build customized software systems, respond to customers more quickly, and expand our footprint in the engineering simulation market.