Efficient Parallel IO For In Situ Data Extracts

OUSD (R&E) MODERNIZATION PRIORITY: Hypersonics TECHNOLOGY AREA(S): Air Platform OBJECTIVE: Develop parallel file readers and writers that work efficiently on HPC systems for in situ data extracts. DESCRIPTION: The goal of this topic is to develop efficient, parallel IO tools for in situ data extracts that work with ParaView Catalyst and VisIt LibSim. The tool should support all VTK (https://www.vtk.org) data types that are supported in ParaView Catalyst and VisIt LibSim. The tool must be open source with an appropriate license to work with both of these libraries (e.g., BSD, Apache) and the data format must be open as well. The data extract output should minimize the number of files generated per output time step and, possibly, aggregate over multiple output time steps. Additionally, the tools would automate the efficiency of the parallel IO, considering significant and varying load imbalance, with as little as possible user parameters. PHASE I: Demonstrate prototype software that combines parallel partitioned VTK data that would normally be written to separate files into a consolidated dataset that minimizes the number of files produced. Characterize the performance improvements of new IO algorithms that can deal with time varying, poorly load-balanced in situ data, through increased IO speed, and reduced number of files produced. PHASE II: Complete in situ workflows that minimize the number of files produced for in situ data extracts per time step for all appropriate VTK data types used in ParaView Catalyst and/or VisIt LibSim. Tools must be in open-source software with open-data formats. Complete effective strategies for efficient parallel IO of in situ data extracts. These strategies should be able to be automatically tuned to specific HPC machine architectures in order to minimize user specified parameters to get the tool to work efficiently. PHASE III DUAL USE APPLICATIONS: In situ use through both ParaView Catalyst and VisIt LibSim are already well established in the CREATE-AVTM program through Helios and Kestrel, respectively. Additionally, in situ use has spread to non-DOD specific simulation codes. See for example, OpenFOAM (https://www.openfoam.com/news/main-news/openfoam-v1806/post-processing). It is expected that this technology will simplify parallel file IO for data extracts for many ParaView Catalyst and VisIt LibSim users. REFERENCES: A. Bauer, H. Abbasi, J. Ahrens, H. Childs, B. Geveci, S. Klasky, K. Moreland, P. O'Leary, V. Vishwanath, B. Whitlock, and E. Bethel, "In Situ Methods, Infrastructures, and Applications on High Performance Computing Platforms," Computer Graphics Forum, vol. 35, no. 3, pp. 577-597, Jun. 2016. Fabian N., Moreland K., Thompson D., Bauer A. C., Marion P., Geveci B., Rasquin M., Jansen K. E.: The paraview coprocessing library: A scalable, general purpose in situ visualization library. In IEEE Symposium on Large-Scale Data Analysis and Visualization (LDAV) 2011 (October 2011), Institute of Electrical and Electronics Engineers, pp. 89 96. S. Herbein, S. McDaniel, N. Podhorszki, J. Logan, S. Klasky, M. Taufer, Performance characterization of irregular I/O at the extreme scale, Parallel Computing, Volume 51, pp 17-35, 2016. Lofstead J., Zheng F., Klasky S., Schwan K.: Adaptable, metadata rich io methods for portable high performance io. In Parallel Distributed Processing, 2009. IPDPS 2009. IEEE International Symposium on (May 2009), pp. 1 10.; 5. Whitlock B., Favre J. M., Meredith J. S.: Parallel in situ coupling of simulation with a fully featured visualization system. In Proceedings of the 11th Eurographics conference on Parallel Graphics and Visualization (2011), Eurographics Association, pp. 101 109. KEYWORDS: HPC, in situ analysis & visualization, parallel IO