The IonOrb code generates a heat map of deposited high energy particles onto the walls of the DIII-D tokamak (blue). This information is crucial for the protection of sensitive equipment and the prevention of excessive impurity generation. (Image: DIII-D National Fusion Facility)
The IonOrb code generates a heat map of deposited high energy particles onto the walls of the DIII-D tokamak (blue). This information is crucial for the protection of sensitive equipment and the prevention of excessive impurity generation. (Image: DIII-D National Fusion Facility)
At the DIII-D National Fusion Facility, researchers study plasmas under fusion-relevant conditions to advance the physics and technology needed for future fusion reactors. This work requires analyzing experimental data quickly and accurately, so scientists can make informed decisions during live operations. To accelerate this process, researchers from DIII-D and ALCF are leveraging integrated workflows and computing capabilities to link experiments directly to DOE supercomputers.
Plasma behavior inside a tokamak is complex and constantly evolving, requiring accurate reconstructions and particle tracking to guide experiments and protect sensitive equipment. Traditional analysis can take hours, delaying feedback and limiting opportunities to adjust experimental parameters in real time. Faster turnaround requires infrastructure that moves data between facilities, runs compute-intensive simulations at scale, and delivers actionable results within minutes.
The team pioneered on-demand remote analysis between DIII-D and ALCF, and is now advancing their work through a DOE Integrated Research Infrastructure (IRI) Pathfinder Project and an ALCC award of computing time at ALCF and NERSC. Central to their approach is a combination of advanced computing capabilities—on-demand queues, direct networking access to compute nodes, and Globus services—that enable fast, secure, and automated analysis. The Consistent Automatic Kinetic Equilibria (CAKE) workflow uses this infrastructure to launch high-fidelity plasma reconstructions in sync with DIII-D’s plasma shots. The IonOrb workflow runs a GPU-accelerated particle-tracking code to simulate high-energy particle trajectories and identify where they deposit energy on the walls of the tokamak vessel. To return results quickly, IonOrb runs on 20 nodes of ALCF’s Polaris (80 GPUs), far beyond the resources available at DIII-D. Both workflows automate data transfers between DIII-D and ALCF to enable remote processing, while on-demand queues provide immediate access to computing resources.
The CAKE workflow now delivers high-fidelity kinetic equilibrium reconstructions closely aligned with DIII-D’s time-intensive experiment cycle, allowing scientists to evaluate plasma performance during ongoing operations. IonOrb is providing power deposition information in minutes, enabling near-real-time analysis, and the identification of potential hot spots caused by the 20MW of neutral beam particle injectors. The team shared their advances at the SC24 conference, where researchers presented a paper at the XLOOP workshop and led a live demonstration of a fusion experiment workflow linking plasma shots to rapid analysis on DOE supercomputers.
By integrating DIII-D experiments with DOE supercomputers, the team is advancing both the speed and fidelity of plasma analysis. This work serves as a model for cross-facility science under DOE’s IRI program and lays the foundation for more compute-intensive analysis at DIII-D, with the potential to bring additional applications into the experiment-time analysis cycle.
Smith, S. P., Z. A. Xing, T. B. Amara, S. S. Denk, E. W. DeShazer, O. Meneghini, T. Neiser, et al. “Expediting Higher Fidelity Plasma State Reconstructions for the DIII-D National Fusion Facility Using Leadership Class Computing Resources,” SC24-W: Workshops of the International Conference for High Performance Computing, Networking, Storage and Analysis (2024), IEEE. https://doi.org/10.1109/SCW63240.2024.00265
Kostuk, M., T. Uram, T. Evans, D. Orlov, M. Papka, and D. Schissel. “Automatic Between-Pulse Analysis of DIII-D Experimental Data Performed Remotely on a Supercomputer at Argonne Leadership Computing Facility,” Fusion Science and Technology (February 2018), Taylor & Francis. https://doi.org/10.1080/15361055.2017.1390388
