Deployment of an Analysis Library for Monitoring of Extreme-scale Global Gyrokinetic Simulations with XGC

Edge gyrokinetic simulations using the X-Point Gyrokinetic Code (XGC) generate petabytes of high-dimensional data critical for understanding magnetically confined plasmas. These simulations run on peta- to exascale HPC systems, where storage limitations and I/O bottlenecks often render traditional post-processing infeasible. To address this, we present a Python-based analysis and visualization library that enables rapid in-situ data access, mesh-aware data structures, and scalable diagnostic routines. Built on the ADIOS2 framework, the library abstracts low-level I/O and constructs physically meaningful representations of the simulation as poloidal cross-sections across the toroidal domain. These topologies provide the basis for computing diagnostic quantities such as flux-surface averages, gradients, and velocity-space moments. The library is designed for extensibility, allowing seamless integration of custom analysis routines from XGC collaborators. We are also developing a web-accessible, live visualization frontend hosted on NERSC’s Spin service, with data streamed from remote simulations via Globus FTP. The system integrates with the EFFIS 2.0 workflow manager to run concurrently with XGC, continuously processing output and updating the dashboard.

This project is supported by the DOE.