Integrating and Characterizing HPC Task Runtime Systems for hybrid AI-HPC workloads

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Merzky, Andre, Titov, Mikhail, Turilli, Matteo, Jha, Shantenu
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866908558362148864
author Merzky, Andre
Titov, Mikhail
Turilli, Matteo
Jha, Shantenu
author_facet Merzky, Andre
Titov, Mikhail
Turilli, Matteo
Jha, Shantenu
contents Scientific workflows increasingly involve both HPC and machine-learning tasks, combining MPI-based simulations, training, and inference in a single execution. Launchers such as Slurm's srun constrain concurrency and throughput, making them unsuitable for dynamic and heterogeneous workloads. We present a performance study of RADICAL-Pilot (RP) integrated with Flux and Dragon, two complementary runtime systems that enable hierarchical resource management and high-throughput function execution. Using synthetic and production-scale workloads on Frontier, we characterize the task execution properties of RP across runtime configurations. RP+Flux sustains up to 930 tasks/s, and RP+Flux+Dragon exceeds 1,500 tasks/s with over 99.6% utilization. In contrast, srun peaks at 152 tasks/s and degrades with scale, with utilization below 50%. For IMPECCABLE.v2 drug discovery campaign, RP+Flux reduces makespan by 30-60% relative to srun/Slurm and increases throughput more than four times on up to 1,024. These results demonstrate hybrid runtime integration in RP as a scalable approach for hybrid AI-HPC workloads.
format Preprint
id arxiv_https___arxiv_org_abs_2509_20819
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Integrating and Characterizing HPC Task Runtime Systems for hybrid AI-HPC workloads
Merzky, Andre
Titov, Mikhail
Turilli, Matteo
Jha, Shantenu
Distributed, Parallel, and Cluster Computing
Scientific workflows increasingly involve both HPC and machine-learning tasks, combining MPI-based simulations, training, and inference in a single execution. Launchers such as Slurm's srun constrain concurrency and throughput, making them unsuitable for dynamic and heterogeneous workloads. We present a performance study of RADICAL-Pilot (RP) integrated with Flux and Dragon, two complementary runtime systems that enable hierarchical resource management and high-throughput function execution. Using synthetic and production-scale workloads on Frontier, we characterize the task execution properties of RP across runtime configurations. RP+Flux sustains up to 930 tasks/s, and RP+Flux+Dragon exceeds 1,500 tasks/s with over 99.6% utilization. In contrast, srun peaks at 152 tasks/s and degrades with scale, with utilization below 50%. For IMPECCABLE.v2 drug discovery campaign, RP+Flux reduces makespan by 30-60% relative to srun/Slurm and increases throughput more than four times on up to 1,024. These results demonstrate hybrid runtime integration in RP as a scalable approach for hybrid AI-HPC workloads.
title Integrating and Characterizing HPC Task Runtime Systems for hybrid AI-HPC workloads
topic Distributed, Parallel, and Cluster Computing
url https://arxiv.org/abs/2509.20819