In addition to an edge server for tasks requiring instantaneous access to high-power compute, the CNMS also possesses traditional high-performance nodes for more traditional data analysis and simulation workflows, including those involving AI/ML. These are managed via an OpenStack cloud platform enabling users to choose different flavors of virtual machines that can come with a range of volume (storage) sizes, numbers of cores, amount of RAM and availability of GPUs. Applications for these nodes include large parallel simulations (taking advantage of multiple CPU cores), GPU-based simulations (E.g., LAMMPS accelerated on GPUs), training of deep learning models, and powerful servers for running Jupyter notebooks for demanding analysis tasks.
The CNMS has four dedicated computational nodes with both CPU and GPU power. Each node contains 2xNVidia A100 GPUs with 80GB of memory for a wide variety of deep learning and accelerated simulation workflow. Additionally, each node AMD EPYC 9654 2.40GHz
Specifications/Capabilities/Extras
• 4 nodes of compute available
• Specifications for each node:
o 2 x NVidia A100 GPUs, 80GB memory each
o 2 x AMD EPYC 2.40GHz CPUs with 96 cores and 384M cache
o 15.36TB Enterprise NVMe Read Intensive AG Drive U.2 Gen4 with carrier
