AI-Enabling Workloads on Large-Scale GPU-Accelerated System: Characterization, Opportunities, and Implications

Production high-performance computing (HPC) systems are adopting and integrating GPUs into their design to accommodate artificial intelligence (AI), machine learning, and data visualization workloads. To aid with the design and operations of new and existing GPU-based large-scale systems, we provide...

Full description

Saved in:
Bibliographic Details
Main Authors: Li, Baolin, Arora, Rohin, Samsi, Siddharth, Patel, Tirthak, Arcand, William, Bestor, David, Byun, Chansup, Roy, Rohan Basu, Bergeron, Bill, Holodnak, John, Houle, Michael, Hubbell, Matthew, Jones, Michael, Kepner, Jeremy, Klein, Anna, Michaleas, Peter, McDonald, Joseph, Milechin, Lauren, Mullen, Julie, Prout, Andrew, Price, Benjamin, Reuther, Albert, Rosa, Antonio, Weiss, Matthew, Yee, Charles, Edelman, Daniel, Vanterpool, Allan, Cheng, Anson, Gadepally, Vijay, Tiwari, Devesh
Format: Conference Proceeding
Language:English
Subjects:
Cluster Characterization
Computer architecture
Data visualization
Deep Learning
GPU Datacenter
Graphics processing units
HPC
Machine learning
Production
System Operation
Systems operation
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Production high-performance computing (HPC) systems are adopting and integrating GPUs into their design to accommodate artificial intelligence (AI), machine learning, and data visualization workloads. To aid with the design and operations of new and existing GPU-based large-scale systems, we provide a detailed characterization of system operations, job characteristics, user behavior, and trends on a contemporary GPU-accelerated production HPC system. Our insights indicate that the pre-mature phases in modern AI workflow take up significant GPU hours while underutilizing GPUs, which opens up the opportunity for a multi-tier system. Finally, we provide various potential recommendations and areas for future investment for system architects, operators, and users.
ISSN:2378-203X
DOI:10.1109/HPCA53966.2022.00093