skip to main content
OSTI.GOV title logo U.S. Department of Energy
Office of Scientific and Technical Information

Title: DRAGON: breaking GPU memory capacity limits with direct NVM access

Abstract

Heterogeneous computing with accelerators is growing in importance in high performance computing (HPC). Recently, application datasets have expanded beyond the memory capacity of these accelerators, and often beyond the capacity of their hosts. Meanwhile, nonvolatile memory (NVM) storage has emerged as a pervasive component in HPC systems because NVM provides massive amounts of memory capacity at affordable cost. Currently, for accelerator applications to use NVM, they must manually orchestrate data movement across multiple memories and this approach only performs well for applications with simple access behaviors. To address this issue, we developed DRAGON, a solution that enables all classes of GP-GPU applications to transparently compute on terabyte datasets residing in NVM. DRAGON leverages the page-faulting mechanism on the recent NVIDIA GPUs by extending capabilities of CUDA Unified Memory (UM). Our experimental results show that DRAGON transparently expands memory capacity and obtain additional speedups via automated I/O and data transfer overlapping.

Authors:
 [1]; ORCiD logo [2]; ORCiD logo [2]; ORCiD logo [2];  [3]
  1. Tokyo Institute of Technology, Japan
  2. ORNL
  3. RIKEN Laboratory
Publication Date:
Research Org.:
Oak Ridge National Lab. (ORNL), Oak Ridge, TN (United States)
Sponsoring Org.:
USDOE
OSTI Identifier:
1489577
DOE Contract Number:  
AC05-00OR22725
Resource Type:
Conference
Resource Relation:
Conference: The International Conference for High Performance Computing, Networking, Storage, and Analysis - Dallas, Texas, United States of America - 11/11/2018 10:00:00 AM-11/16/2018 10:00:00 AM
Country of Publication:
United States
Language:
English

Citation Formats

Markthub, Pak, Belviranli, Mehmet E., Lee, Seyong, Vetter, Jeffrey S., and Matsuoka, Satoshi. DRAGON: breaking GPU memory capacity limits with direct NVM access. United States: N. p., 2018. Web. doi:10.1109/SC.2018.00035.
Markthub, Pak, Belviranli, Mehmet E., Lee, Seyong, Vetter, Jeffrey S., & Matsuoka, Satoshi. DRAGON: breaking GPU memory capacity limits with direct NVM access. United States. doi:10.1109/SC.2018.00035.
Markthub, Pak, Belviranli, Mehmet E., Lee, Seyong, Vetter, Jeffrey S., and Matsuoka, Satoshi. Thu . "DRAGON: breaking GPU memory capacity limits with direct NVM access". United States. doi:10.1109/SC.2018.00035. https://www.osti.gov/servlets/purl/1489577.
@article{osti_1489577,
title = {DRAGON: breaking GPU memory capacity limits with direct NVM access},
author = {Markthub, Pak and Belviranli, Mehmet E. and Lee, Seyong and Vetter, Jeffrey S. and Matsuoka, Satoshi},
abstractNote = {Heterogeneous computing with accelerators is growing in importance in high performance computing (HPC). Recently, application datasets have expanded beyond the memory capacity of these accelerators, and often beyond the capacity of their hosts. Meanwhile, nonvolatile memory (NVM) storage has emerged as a pervasive component in HPC systems because NVM provides massive amounts of memory capacity at affordable cost. Currently, for accelerator applications to use NVM, they must manually orchestrate data movement across multiple memories and this approach only performs well for applications with simple access behaviors. To address this issue, we developed DRAGON, a solution that enables all classes of GP-GPU applications to transparently compute on terabyte datasets residing in NVM. DRAGON leverages the page-faulting mechanism on the recent NVIDIA GPUs by extending capabilities of CUDA Unified Memory (UM). Our experimental results show that DRAGON transparently expands memory capacity and obtain additional speedups via automated I/O and data transfer overlapping.},
doi = {10.1109/SC.2018.00035},
journal = {},
number = ,
volume = ,
place = {United States},
year = {2018},
month = {11}
}

Conference:
Other availability
Please see Document Availability for additional information on obtaining the full-text document. Library patrons may search WorldCat to identify libraries that hold this conference proceeding.

Save / Share: