GTC 2020: Building Optimized, Low-Cost, Scalable Health-Care Enterprise Deep Learning Services Platform with NVIDIA TensorRT Inference Server and Kubernetes
After clicking “Watch Now” you will be prompted to login or join.
Click “Watch Now” to login or join the NVIDIA Developer Program.
Building Optimized, Low-Cost, Scalable Health-Care Enterprise Deep Learning Services Platform with NVIDIA TensorRT Inference Server and Kubernetes
Galina Grunin , Optum | Dima Rekesh, Optum Tech, United Health Group
The HPC system "Perlmutter" will the first GPU-accelerated production system at the U.S. Department of Energy's National Energy Research Scientific Computing Center (NERSC) when it is deployed in 2021. We'll explain how, to enable its users to prepare their applications for Perlmutter, NERSC recently integrated 18 GPU-accelerated compute nodes into its current production system, the Knights Landing-based Cray system "Cori." These nodes' primary purpose is application development and profiling for GPU acceleration, as part of the NERSC Exascale Science Application Program (NESAP). Despite significant differences in hardware from the rest of the Cori system, the GPU nodes have been configured such that, from both the user's and the administrator's perspective, they are seamlessly integrated into Cori. This integration has streamlined access for hundreds of NESAP users to these nodes, and facilitates a diverse workload of NESAP applications spanning simulation, data-intensive workloads, and machine learning.