Comparing single-node and multi-node performance of an important fusion HPC code benchmark

TL;DR

This paper demonstrates that high-performance fusion simulations can be efficiently run on single-node multi-GPU systems, outperforming traditional multi-node HPC setups for certain benchmarks, with implications for resource optimization.

Contribution

It provides a detailed performance analysis of the CGYRO fusion simulation on single-node GPU systems, showing potential for replacing multi-node setups for specific problem sizes.

Findings

Single-node GPU setups can outperform multi-node HPC for certain benchmarks.

Larger problems still require multi-node HPC due to GPU memory limits.

Upcoming hardware like NVSWITCH could further enhance single-node GPU performance.

Abstract

Fusion simulations have traditionally required the use of leadership scale High Performance Computing (HPC) resources in order to produce advances in physics. The impressive improvements in compute and memory capacity of many-GPU compute nodes are now allowing for some problems that once required a multi-node setup to be also solvable on a single node. When possible, the increased interconnect bandwidth can result in order of magnitude higher science throughput, especially for communication-heavy applications. In this paper we analyze the performance of the fusion simulation tool CGYRO, an Eulerian gyrokinetic turbulence solver designed and optimized for collisional, electromagnetic, multiscale simulation, which is widely used in the fusion research community. Due to the nature of the problem, the application has to work on a large multi-dimensional computational mesh as a whole, requiring frequent exchange of large amounts of data between the compute processes. In particular, we show that the average-scale nl03 benchmark CGYRO simulation can be run at an acceptable speed on a single Google Cloud instance with 16 A100 GPUs, outperforming 8 NERSC Perlmutter Phase1 nodes, 16 ORNL Summit nodes and 256 NERSC Cori nodes. Moving from a multi-node to a single-node GPU setup we get comparable simulation times using less than half the number of GPUs. Larger benchmark problems, however, still require a multi-node HPC setup due to GPU memory capacity needs, since at the time of writing no vendor offers nodes with a sufficient GPU memory setup. The upcoming external NVSWITCH does however promise to deliver an almost equivalent solution for up to 256 NVIDIA GPUs.