research-article

Public Access

DPU-Bench: A Micro-Benchmark Suite to Measure Offload Efficiency Of SmartNICs

Authors:

Benjamin Michalowicz,

Kaushik Kandadi Suresh,

Hari Subramoni,

Dhabaleswar Panda,

Steve PooleAuthors Info & Claims

PEARC '23: Practice and Experience in Advanced Research Computing 2023: Computing for the Common Good

Pages 94 - 101

https://doi.org/10.1145/3569951.3593595

Published: 10 September 2023 Publication History

All formats PDF

Abstract

Smart Network Interface Cards (SmartNIC) have experienced massive growth in popularity over the last few years such as the NVIDIA BlueField-2 Data Processing Unit (DPU). Being equipped with their own set of cores and memory allows them to perform actions beyond a regular NIC, and HPC researchers are designing new ways to use them. For example, offloading communication to one enables the CPU "host" to perform more computationally heavy tasks. However, one question remains: How much of that work can be distributed among processes placed on the SmartNIC before facing performance degradation? We present DPU-Bench: A low-level micro-benchmark suite using IB-Verbs primitives to enable HPC users to examine the number of processes to be placed on one or more SmartNICs in order to efficiently offload a given communication pattern. We examine direct algorithms in this paper at a medium scale with different work assignment mechanisms and give insights into the trends found with varying numbers of worker processes and message sizes.

References

[1]

Argonne National Laboratory 2022. MPICH. https://www.mpich.org/.

[2]

Dotan Barak. 2014. Verbs Programming Tutorial. https://www.cs.mtsu.edu/ waderholdt/6430/papers/ibverbs.pdf

[3]

Mohammadreza Bayatpour, Nick Sarkauskas, Hari Subramoni, Jahanzeb Maqbool Hashmi, and Dhabaleswar K. Panda. 2021. BluesMPI: Efficient MPI Non-blocking Alltoall Offloading Designs on Modern BlueField Smart NICs. In High Performance Computing, Bradford L. Chamberlain, Ana-Lucia Varbanescu, Hatem Ltaief, and Piotr Luszczek (Eds.). Springer International Publishing, Cham, 18–37.

[4]

Idan Burstein. 2021. Nvidia Data Center Processing Unit (DPU) Architecture. In 2021 IEEE Hot Chips 33 Symposium (HCS). 1–20. https://doi.org/10.1109/HCS52781.2021.9567066

[5]

High Performance Compute Availability Group 2022. OpenHPCA Benchmark Suite. https://github.com/openucx/openhpca.

[6]

Intel 2022. Intel MPI Benchmarks. https://www.intel.com/content/www/us/en/developer/articles/technical/intel-mpi-benchmarks.html.

[7]

Arpan Jain, Nawras Alnaasan, Aamir Shafi, Hari Subramoni, and Dhabaleswar K Panda. 2021. Accelerating CPU-based Distributed DNN Training on Modern HPC Clusters using BlueField-2 DPUs. In 2021 IEEE Symposium on High-Performance Interconnects (HOTI). 17–24. https://doi.org/10.1109/HOTI52880.2021.00017

[8]

Pouya Kousha, Kamal Raj Sankarapandian Dayala Ganesh Ram, Mansa Kedia, Hari Subramoni, Arpan Jain, Aamir Shafi, Dhabaleswar Panda, Trey Dockendorf, Heechang Na, and Karen Tomko. 2021. INAM: Cross-Stack Profiling and Analysis of Communication in MPI-Based Applications. In Practice and Experience in Advanced Research Computing (Boston, MA, USA) (PEARC ’21). Association for Computing Machinery, New York, NY, USA, Article 14, 11 pages. https://doi.org/10.1145/3437359.3465582

Digital Library

[9]

Patrick MacArthur, Qian Liu, Robert D. Russell, Fabrice Mizero, Malathi Veeraraghavan, and John M. Dennis. 2017. An Integrated Tutorial on InfiniBand, Verbs, and MPI. IEEE Communications Surveys & Tutorials 19, 4 (2017), 2894–2926. https://doi.org/10.1109/COMST.2017.2746083

[10]

Message Passing Interface Forum. 2021. MPI: A Message-Passing Interface Standard Version 4.0. https://www.mpi-forum.org/docs/mpi-4.0/mpi40-report.pdf

[11]

Network-Based Computing Laboratory 2022. MVAPICH: MPI over InfiniBand, 10GigE/iWARP and RoCE. http://mvapich.cse.ohio-state.edu/.

[12]

Network-Based Computing Laboratory 2022. OSU Microbenchmarks. http://mvapich.cse.ohio-state.edu/benchmarks/.

[13]

Nick Sarkauskas, Mohammadreza Bayatpour, Tu Tran, Bharath Ramesh, Hari Subramoni, and Dhabaleswar K. Panda. 2021. Large-Message Nonblocking MPI_Iallgather and MPI Ibcast Offload via BlueField-2 DPU. In 2021 IEEE 28th International Conference on High Performance Computing, Data, and Analytics (HiPC). 388–393. https://doi.org/10.1109/HiPC53243.2021.00054

[14]

Sameer S. Shende and Allen D. Malony. 2006. The Tau Parallel Performance System. Int. J. High Perform. Comput. Appl. 20, 2 (may 2006), 287–311. https://doi.org/10.1177/1094342006064482

Digital Library

[15]

Dan Terpstra, Heike Jagode, Haihang You, and Jack Dongarra. 2010. Collecting Performance Data with PAPI-C. In Tools for High Performance Computing 2009, Matthias S. Müller, Michael M. Resch, Alexander Schulz, and Wolfgang E. Nagel (Eds.). Springer Berlin Heidelberg, Berlin, Heidelberg, 157–173. https://doi.org/10.1007/978-3-642-11261-4_11

[16]

Top500 2022. Top500. https://www.top500.org/lists/top500/2022/11/.

Cited By

Basu SNadig D(2024)Offloading NVMe over Fabrics (NVMe-oF) to SmartNICs on an at-scale Distributed Testbed2024 IEEE 10th International Conference on Network Softwarization (NetSoft)10.1109/NetSoft60951.2024.10588915(316-318)Online publication date: 24-Jun-2024
https://doi.org/10.1109/NetSoft60951.2024.10588915
Michalowicz BSuresh KSubramoni HPanda DPoole S(2023)Battle of the BlueFields: An In-Depth Comparison of the BlueField-2 and BlueField-3 SmartNICs2023 IEEE Symposium on High-Performance Interconnects (HOTI)10.1109/HOTI59126.2023.00020(41-48)Online publication date: Aug-2023
https://doi.org/10.1109/HOTI59126.2023.00020

Recommendations

An OpenCL micro-benchmark suite for GPUs and CPUs

Open computing language (OpenCL) is a new industry standard for task-parallel and data-parallel heterogeneous computing on a variety of modern CPUs, GPUs, DSPs, and other microprocessor designs. OpenCL is vendor independent and hence not specialized for ...
CUDA acceleration for Xen virtual machines in infiniband clusters with rCUDA
PPoPP '16

Many data centers currently use virtual machines (VMs) to achieve a more efficient usage of hardware resources. However, current virtualization solutions, such as Xen, do not easily provide graphics processing unit (GPU) accelerators to applications ...
CUDA acceleration for Xen virtual machines in infiniband clusters with rCUDA
PPoPP '16: Proceedings of the 21st ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming

Many data centers currently use virtual machines (VMs) to achieve a more efficient usage of hardware resources. However, current virtualization solutions, such as Xen, do not easily provide graphics processing unit (GPU) accelerators to applications ...

Comments

Information & Contributors

Information

Published In

cover image ACM Conferences

PEARC '23: Practice and Experience in Advanced Research Computing 2023: Computing for the Common Good

July 2023

519 pages

ISBN:9781450399852

DOI:10.1145/3569951

Editors:
Robert Sinkovits
San Diego Supercomputer Center
,
Alana Romanella
University of Colorado Boulder
,
Shelley Knuth
University of Colorado Boulder
,
Ken Hackworth
Pittsburgh Supercomputing Center
,
Jeff Pummill
University of Arkansas

Copyright © 2023 ACM.

Publication rights licensed to ACM. ACM acknowledges that this contribution was authored or co-authored by an employee, contractor or affiliate of the United States government. As such, the Government retains a nonexclusive, royalty-free right to publish or reproduce this article, or to allow others to do so, for Government purposes only.

Sponsors

Publisher

Association for Computing Machinery

New York, NY, United States

Publication History

Published: 10 September 2023

Permissions

Request permissions for this article.

Request Permissions

Check for updates

Author Tags

Qualifiers

Research-article
Research
Refereed limited

Funding Sources

Conference

PEARC '23

Sponsor:

PEARC '23: Practice and Experience in Advanced Research Computing

July 23 - 27, 2023

OR, Portland, USA

Acceptance Rates

Overall Acceptance Rate 133 of 202 submissions, 66%

Contributors

Other Metrics

View Article Metrics

Bibliometrics & Citations

Bibliometrics

Article Metrics

2
Total Citations
View Citations
331
Total Downloads

Downloads (Last 12 months)293
Downloads (Last 6 weeks)37

Reflects downloads up to 16 Oct 2024

Other Metrics

View Author Metrics

Citations

Cited By

Basu SNadig D(2024)Offloading NVMe over Fabrics (NVMe-oF) to SmartNICs on an at-scale Distributed Testbed2024 IEEE 10th International Conference on Network Softwarization (NetSoft)10.1109/NetSoft60951.2024.10588915(316-318)Online publication date: 24-Jun-2024
https://doi.org/10.1109/NetSoft60951.2024.10588915
Michalowicz BSuresh KSubramoni HPanda DPoole S(2023)Battle of the BlueFields: An In-Depth Comparison of the BlueField-2 and BlueField-3 SmartNICs2023 IEEE Symposium on High-Performance Interconnects (HOTI)10.1109/HOTI59126.2023.00020(41-48)Online publication date: Aug-2023
https://doi.org/10.1109/HOTI59126.2023.00020

View Options

View options

PDF

View or Download as a PDF file.

eReader

View online with eReader.

HTML Format

View this article in HTML Format.

Get Access

Login options

Check if you have access through your login credentials or your institution to get full access on this article.

Full Access

Get this Publication

Media

Figures

Other

Tables

View Table of Contents