Today Intelligent Light announced breakthrough CFD scalability running the AVF-Leslie combustion simulation code on up to 64,000 cores on NERSC supercomputers. As part of a collaboration with scalable solver developers at Georgia Tech and HPC experts at LBNL, the project is bringing together Intelligent Light’s expertise in computer science, software and hardware architecture, solver integration, data management, and the practical application of CFD to deliver scalable analysis methods and in situ infrastructure for extreme scale knowledge discovery.
Intelligent Light has been known for years for our post-processing and visualization technology. As the computing landscapes shifts to high-performance clusters, integrating post-processing with CFD solvers presents an opportunity to create a truly scalable workflow.”
High performance computing is a necessity for the pursuit of high-fidelity, time-accurate simulations of sophisticated physics. HPC makes it possible to run these highly detailed simulations and deliver results in reasonable timeframes. When running simulations using thousands of cores however, the time to write, re-read and post-process the resulting files using traditional volume-based post-processing is impractical or impossible. When results are not reviewed and desired simulation runs not performed due to these limitations, the cost is wasted computing resources and lost science. In situ methods enable analysis of full spatiotemporal resolution data while it is still resident in memory, thereby avoiding the costs associated with writing very large data files to persistent storage for subsequent, post hoc analysis.
Scalable CFD analysis with extreme scale computing
For the AVF-Leslie code, a derivative of Georgia Tech’s Leslie3D solver code, breakthrough scalability is being achieved when running on up to 64,000 cores and the code has been instrumented with VisIt/Libsim to enable in situ extraction of surfaces of interest. Previously, the code has been used for combustion simulations running up to 5,000 cores. The extracted surfaces are output to compact XDB files for secondary processing using FieldView, Intelligent Light’s highly efficient post-processing tool that is a mainstay of CFD analysis. XDBs retain full numerical fidelity, enable both automated report generation and interactive exploration and can be used for archives. Already, combustion researchers are simulating at this extreme scale and learning about phenomena never before possible to explore either numerically or experimentally.
Intelligent Light has been known for years for our post-processing and visualization technology. As the computing landscapes shifts to high-performance clusters, integrating post-processing with CFD solvers presents an opportunity to create a truly scalable workflow,” said Steve M. Legensky, General Manager and Founder at Intelligent Light. “Although we have run the VisIt code at up to 98,000 cores on the LLNL BlueGene/Q systems, we are seeing the challenges that arise when integrating with a sophisticated physics code like AVF-Leslie. This project, as well as others we have executed for DOE and DoD are helping us to understand the issues that affect both post-processing and solver code performance so that we can help our customers be successful in the HPC world.”
DOE taps Intelligent Light expertise in pursuit of extreme scale coherency and production quality software for real-world science
The Department of Energy (DOE) selected Intelligent Light as part of a team led by Lawrence Berkeley National Laboratory — a team that also includes Kitware, Georgia Tech and Argonne National Laboratory — to address the challenges of extreme-scale computing and integrating CFD solvers with in situ methods. Software tools for integrating solvers with in situ methods at extreme scale must maintain coherency across tens to hundreds of thousands of processor cores and be production quality to produce useful scientific results.
Today we see widening gaps between compute performance and I/O capability and in situ analysis is a key part of the solution. As we move toward the exascale regime, we will see 3 orders of magnitude increase in FLOPs performance while at the same time seeing only 2-3 times more I/O performance,” says Wes Bethel, Senior Computer Scientist at Lawrence Berkeley National Laboratory. “Next generation workflows must address this discrepancy while delivering ultra-scalable performance for applications and Intelligent Light is among the organizations that are developing the proven, production quality software that will be required to produce successful science from these machines.”
DOE has assembled an exclusive team to develop the next generation methods and tools for in situ workflows to be used in a wide range of HPC-based scientific applications. Libsim is a key interface for in situ applications. As Libsim is tightly coupled to the solver, Intelligent Light is working with solvers to integrate this interface. Intelligent Light is a leading developer and maintainer of the open-source VisIt application and Libsim, both developed by the DOE.
XDB – Extract database files provide essential capability for interrogation
The use of extracts permits post hoc interactive exploration using standard tools without requiring the user to know what they want to see in advance. XDBs files are 10-1000 times smaller than solution files and are computationally efficient to create and save. CFD users can utilize automation to analyze large volumes of data, apply new generation techniques to identify important features from across vast datasets, and maintain the ability to explore solutions interactively using FieldView – the highly efficient, user-centric post-processing product long a favorite tool of CFD practitioners across industry and research around the world.
Leading the way forward with HPC
In working on research and production projects at the extreme scale, Intelligent Light is developing leading edge expertise and experience solving the challenges that occur when in situ methods are deployed in real world applications at the extreme scale. By understanding and solving the scalability and workflow issues at 64,000 cores and beyond, the use of HPC and in situ will be accelerated for all Intelligent Light customers.
To support the development and deployment of in situ methods, an SC15 workshop has been organized. Research results are being periodically presented as Intelligent Light’s research and development progress. The combustion study results were recently presented at the 27th International Conference on Parallel Computational Fluid Dynamics and presentations on related research are planned for AIAA SciTech in January, 2016.
This work is supported by the U.S. Department of Energy, Office of Science, Office of Advanced Scientific Computing Research under Award Number DE-SC0012449.