research-article

Snowflakes at the Edge: A Study of Variability among NVIDIA Jetson AGX Xavier Boards

Authors:

Hazem A. Abdelhafez,

Karthik Pattabiraman,

Matei RipeanuAuthors Info & Claims

EdgeSys '21: Proceedings of the 4th International Workshop on Edge Systems, Analytics and Networking

Pages 1 - 6

https://doi.org/10.1145/3434770.3459729

Published: 26 April 2021 Publication History

Abstract

While applications deployed at the edge often rely on performance stability (or, at a minimum, on a predictable level of performance), variability at the edge remains a real problem [4]. This study uncovers a surprising source of variability: intrinsic variability (in performance and power consumption) among edge platforms that are nominally identical. We focus on a popular platform designed for edge applications, the NVIDIA Jetson AGX, and aim to answer the following high-level questions through rigorous statistical analysis: (i) are the edge devices in our study statistically different from each other in terms of applications' runtime performance and power draw (although they are sold under the same product model and family)?, (ii) if the differences between these edge devices are statistically significant, what is the magnitude of these differences?, and (iii) do these differences matter from the application's perspective?

References

[1]

J. Cohen. 2013. Statistical power analysis for the behavioral sciences. A. Press.

[2]

A. Paszke et al. 2019. PyTorch: An Imperative Style, High-Performance Deep Learning Library. arXiv:1912.01703 [cs.LG]

[3]

B. Gaudette et al. 2016. Improving smartphone user experience by balancing performance and energy with probabilistic QoS guarantee. In 2016 IEEE International Symposium on High Performance Computer Architecture (HPCA). 52--63.

[4]

B. Gaudette et al. 2019. Optimizing User Satisfaction of Mobile Workloads Subject to Various Sources of Uncertainties. IEEE Transactions on Mobile Computing 18, 12 (2019), 2941--2953. https://doi.org/10.1109/TMC.2018.2883619

[5]

B. Kocoloski et al. 2018. Varbench: an Experimental Framework to Measure and Characterize Performance Variability. In Proceedings of the 47th International Conference on Parallel Processing. 1--10.

Digital Library

[6]

C Wu et al. 2019. Machine learning at facebook: Understanding inference at the edge. In 2019 IEEE International Symposium on High Performance Computer Architecture (HPCA). IEEE, 331--344.

[7]

F. Scholz et al. 1987. K-Sample Anderson-Darling Tests. J. Amer. Statist. Assoc. 82, 399 (1987), 918--924. https://doi.org/10.1080/01621459.1987.10478517

[8]

H. Weisbach et al. 2018. Hardware Performance Variation: A Comparative Study Using Lightweight Kernels. In High Performance Computing. Springer International Publishing, Cham, 246--265.

[9]

J. Algina et al. 2005. An alternative to Cohen's standardized mean difference effect size: a robust parameter and confidence interval in the two independent groups case. Psychological methods 10, 3 (2005), 317.

[10]

J. McCullough et al. 2011. Evaluating the effectiveness of model-based power characterization. In USENIX Annual Technical Conf, Vol. 20.

Digital Library

[11]

L. Wanner et al. 2013. Hardware Variability-Aware Duty Cycling for Embedded Sensors. IEEE Transactions on Very Large Scale Integration (VLSI) Systems 21, 6 (2013), 1000--1012. https://doi.org/10.1109/TVLSI.2012.2203325

Digital Library

[12]

M. G. Sarwar Murshed et al. 2020. Machine Learning at the Network Edge: A Survey. arXiv:1908.00080 [cs.LG]

[13]

N. Razali et al. 2011. Power comparisons of shapiro-wilk, kolmogorov-smirnov, lilliefors and anderson-darling tests. Journal of statistical modeling and analytics 2, 1 (2011), 21--33.

[14]

P. Gupta et al. 2012. Underdesigned and opportunistic computing in presence of hardware variability. IEEE Transactions on Computer-Aided Design of integrated circuits and systems 32, 1 (2012), 8--23.

Digital Library

[15]

S. Marcel et al. 2010. Torchvision the Machine-Vision Package of Torch. In Proceedings of the 18th ACM International Conference on Multimedia (Firenze, Italy) (MM '10). Association for Computing Machinery, New York, NY, USA, 1485--1488. https://doi.org/10.1145/1873951.1874254

Digital Library

[16]

J. Hodges. 1958. The significance probability of the Smirnov two-sample test. Arkiv för Matematik 3, 5 (1958), 469--486.

[17]

J. Li. 2016. Effect size measures in a two-independent-samples case with non-normal and nonhomogeneous data. Behavior research methods 48, 4 (2016), 1560--1574.

[18]

NVIDIA 2020. Jetson AGX Xavier Series: Thermal Design Guide. NVIDIA. Retrieved February, 2021 from https://tinyurl.com/r7zeehya

[19]

Texas Instruments 2016. INA3221 Power Monitors. Texas Instruments. Retrieved January, 2021 from https://www.ti.com/product/INA3221

[20]

The PyTorch Team. 2021. CUDA Timing Events. NVIDIA. Retrieved January, 2021 from https://pytorch.org/docs/stable/cuda.html#torch.cuda.Event

Cited By

Aldahoud AFezari MAl-Dahoud AAqel D(2024)A Systematic Review on Applying Machine Learning and Deep Learning on SBCsWSEAS TRANSACTIONS ON INFORMATION SCIENCE AND APPLICATIONS10.37394/23209.2024.21.2621(272-281)Online publication date: 15-May-2024
https://doi.org/10.37394/23209.2024.21.26
Huang YLin ZYao TMo CShang XCui LYang Y(2024)Mobility-Aware Seamless Virtual Function Migration in Deviceless Edge Computing EnvironmentsIEEE Transactions on Mobile Computing10.1109/TMC.2023.334396923:7(7999-8014)Online publication date: Jul-2024
https://doi.org/10.1109/TMC.2023.3343969
Huang YYao TLin ZShang XYuan YCui LYang Y(2024)Efficient Service Function Chain Placement over Heterogeneous Devices in Deviceless Edge Computing EnvironmentsIEEE Transactions on Computers10.1109/TC.2024.3475590(1-14)Online publication date: 2024
https://doi.org/10.1109/TC.2024.3475590
Show More Cited By

Index Terms

Snowflakes at the Edge: A Study of Variability among NVIDIA Jetson AGX Xavier Boards
1. Computer systems organization
  1. Embedded and cyber-physical systems
    1. Embedded systems

Recommendations

Hot Under the Hood: An Analysis of Ambient Temperature Impact on Heterogeneous Edge Platforms
EdgeSys '23: Proceedings of the 6th International Workshop on Edge Systems, Analytics and Networking

Applications deployed at the edge are often subject to critical Quality of Service (QoS) objectives, such as meeting deadlines while optimizing for energy consumption. To design and operate middleware that satisfies these QoS objectives, it is crucial ...
EC-ECC: Accelerating Elliptic Curve Cryptography for Edge Computing on Embedded GPU TX2
Driven by artificial intelligence and computer vision industries, Graphics Processing Units (GPUs) are now rapidly achieving extraordinary computing power. In particular, the NVIDIA Tegra K1/X1/X2 embedded GPU platforms, which are also treated as edge ...
Tango of edge and cloud execution for reliability
MECC '19: Proceedings of the 4th Workshop on Middleware for Edge Clouds & Cloudlets

The increasing cost of cloud services and the need for decentralization of servers has led to a rise of interest in Edge computing. Edge computing brings the servers closer to the end users which helps in reducing network latency. However, the presence ...

Comments

Information & Contributors

Information

Published In

cover image ACM Conferences

EdgeSys '21: Proceedings of the 4th International Workshop on Edge Systems, Analytics and Networking

April 2021

84 pages

ISBN:9781450382915

DOI:10.1145/3434770

General Chairs:
Aaron Ding
TU Delft, The Netherlands
,
Richard Mortier
University of Cambridge, UK

Copyright © 2021 ACM.

Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. Copyrights for components of this work owned by others than the author(s) must be honored. Abstracting with credit is permitted. To copy otherwise, or republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee. Request permissions from [email protected].

Sponsors

SIGOPS: ACM Special Interest Group on Operating Systems

Publisher

Association for Computing Machinery

New York, NY, United States

Publication History

Published: 26 April 2021

Permissions

Request permissions for this article.

Request Permissions

Check for updates

Author Tags

Qualifiers

Research-article
Research
Refereed limited

Conference

EuroSys '21

Sponsor:

SIGOPS

EuroSys '21: Sixteenth European Conference on Computer Systems

April 26, 2021

Online, United Kingdom

Acceptance Rates

Overall Acceptance Rate 10 of 23 submissions, 43%

Upcoming Conference

EuroSys '25

Sponsor:
sigops

Twentieth European Conference on Computer Systems

March 30 - April 3, 2025

Rotterdam , Netherlands

Contributors

Other Metrics

View Article Metrics

Bibliometrics & Citations

Bibliometrics

Article Metrics

13
Total Citations
View Citations
459
Total Downloads

Downloads (Last 12 months)62
Downloads (Last 6 weeks)5

Reflects downloads up to 08 Feb 2025

Other Metrics

View Author Metrics

Citations

Cited By

Aldahoud AFezari MAl-Dahoud AAqel D(2024)A Systematic Review on Applying Machine Learning and Deep Learning on SBCsWSEAS TRANSACTIONS ON INFORMATION SCIENCE AND APPLICATIONS10.37394/23209.2024.21.2621(272-281)Online publication date: 15-May-2024
https://doi.org/10.37394/23209.2024.21.26
Huang YLin ZYao TMo CShang XCui LYang Y(2024)Mobility-Aware Seamless Virtual Function Migration in Deviceless Edge Computing EnvironmentsIEEE Transactions on Mobile Computing10.1109/TMC.2023.334396923:7(7999-8014)Online publication date: Jul-2024
https://doi.org/10.1109/TMC.2023.3343969
Huang YYao TLin ZShang XYuan YCui LYang Y(2024)Efficient Service Function Chain Placement over Heterogeneous Devices in Deviceless Edge Computing EnvironmentsIEEE Transactions on Computers10.1109/TC.2024.3475590(1-14)Online publication date: 2024
https://doi.org/10.1109/TC.2024.3475590
Curry JLouri AKaranth ABunescu R(2024)PCM Enabled Low-Power Photonic Accelerator for Inference and Training on Edge Devices2024 IEEE International Parallel and Distributed Processing Symposium Workshops (IPDPSW)10.1109/IPDPSW63119.2024.00118(600-607)Online publication date: 27-May-2024
https://doi.org/10.1109/IPDPSW63119.2024.00118
Ahmadi AAbdelhafez HPattabiraman KRipeanu MSha KBanerjee SChen J(2023)EdgeEngine: A Thermal-Aware Optimization Framework for Edge InferenceProceedings of the Eighth ACM/IEEE Symposium on Edge Computing10.1145/3583740.3626616(67-79)Online publication date: 6-Dec-2023
https://dl.acm.org/doi/10.1145/3583740.3626616
Diéguez ALópez M(2023)Performance Tuning for GPU-Embedded Systems: Machine-Learning-Based and Analytical Model-Driven Tuning Methodologies2023 IEEE 35th International Symposium on Computer Architecture and High Performance Computing (SBAC-PAD)10.1109/SBAC-PAD59825.2023.00022(129-140)Online publication date: 17-Oct-2023
https://doi.org/10.1109/SBAC-PAD59825.2023.00022
Amendola JDayal ACenkeramaddi LJha A(2023)Edge-Distributed Fusion of Camera-LiDAR for Robust Moving Object LocalizationIEEE Access10.1109/ACCESS.2023.329521211(73583-73598)Online publication date: 2023
https://doi.org/10.1109/ACCESS.2023.3295212
Li CBian SWu TDonovan RLi B(2022)Affordable Artificial Intelligence-Assisted Machine Supervision System for the Small and Medium-Sized ManufacturersSensors10.3390/s2216624622:16(6246)Online publication date: 19-Aug-2022
https://doi.org/10.3390/s22166246
S.K PKesanapalli SSimmhan Y(2022)Characterizing the Performance of Accelerated Jetson Edge Devices for Training Deep Learning ModelsProceedings of the ACM on Measurement and Analysis of Computing Systems10.1145/35706046:3(1-26)Online publication date: 8-Dec-2022
https://dl.acm.org/doi/10.1145/3570604
Huang YLin ZYao TShang XCui LHuang J(2022)Mobility-aware Seamless Virtual Function Migration in Deviceless Edge Computing Environments2022 IEEE 42nd International Conference on Distributed Computing Systems (ICDCS)10.1109/ICDCS54860.2022.00050(447-457)Online publication date: Jul-2022
https://doi.org/10.1109/ICDCS54860.2022.00050
Show More Cited By

View Options

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

View options

PDF

View or Download as a PDF file.

eReader

View online with eReader.

Figures

Tables

Media

View Table of Conten