Memory Reference Analysis and Implications for Executing AI Workloads in Mobile Systems

Seokmin Kwon; Hyokyung Bahn

doi:10.1109/IEIT59852.2023.10335577

Memory Reference Analysis and Implications for Executing AI Workloads in Mobile Systems

Seokmin Kwon, Hyokyung Bahn

Department of Computer Science & Engineering

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

Abstract

Recently, mobile applications that internally execute artificial intelligence workloads are increasing. In mobile environments, performance degradation due to memory thrashing may occur while artificial intelligence workloads perform the training of large data set due to limitations in memory capacity. In this paper, we analyze the memory reference characteristics of artificial intelligence workloads, and observe that artificial intelligence workloads can cause performance degradation by generating a lot of I/Os (Inputs / Outputs) to NAND flash storage in mobile systems due to weak temporal locality and irregular popularity bias in memory write operations. Based on this observation, we discuss system architectures that can efficiently execute artificial intelligence workloads in mobile systems. Specifically, we adopt small persistent memory as write accelerator and show how efficiently the memory write operation of mobile systems can be managed. Simulation experiments show that the proposed system architecture can improve I/O time significantly compared to existing mobile systems.

Original language	English
Title of host publication	Proceedings - IEIT 2023
Subtitle of host publication	2023 International Conference on Electrical and Information Technology
Publisher	Institute of Electrical and Electronics Engineers Inc.
Pages	281-285
Number of pages	5
ISBN (Electronic)	9798350327298
DOIs	https://doi.org/10.1109/IEIT59852.2023.10335577
State	Published - 2023
Event	2023 International Conference on Electrical and Information Technology, IEIT 2023 - Malang, Indonesia Duration: 14 Sep 2023 → 15 Sep 2023

Publication series

Name	Proceedings - IEIT 2023: 2023 International Conference on Electrical and Information Technology

Conference

Conference	2023 International Conference on Electrical and Information Technology, IEIT 2023
Country/Territory	Indonesia
City	Malang
Period	14/09/23 → 15/09/23

Bibliographical note

Publisher Copyright:
© 2023 IEEE.

Keywords

AI workload
Mobile system
machine learning
memory reference
write acceleration

Access to Document

10.1109/IEIT59852.2023.10335577

Cite this

Kwon, S., & Bahn, H. (2023). Memory Reference Analysis and Implications for Executing AI Workloads in Mobile Systems. In Proceedings - IEIT 2023: 2023 International Conference on Electrical and Information Technology (pp. 281-285). (Proceedings - IEIT 2023: 2023 International Conference on Electrical and Information Technology). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/IEIT59852.2023.10335577

Kwon, Seokmin ; Bahn, Hyokyung. / Memory Reference Analysis and Implications for Executing AI Workloads in Mobile Systems. Proceedings - IEIT 2023: 2023 International Conference on Electrical and Information Technology. Institute of Electrical and Electronics Engineers Inc., 2023. pp. 281-285 (Proceedings - IEIT 2023: 2023 International Conference on Electrical and Information Technology).

@inproceedings{3c88eba264054963af100b87e2bd94ad,

title = "Memory Reference Analysis and Implications for Executing AI Workloads in Mobile Systems",

abstract = "Recently, mobile applications that internally execute artificial intelligence workloads are increasing. In mobile environments, performance degradation due to memory thrashing may occur while artificial intelligence workloads perform the training of large data set due to limitations in memory capacity. In this paper, we analyze the memory reference characteristics of artificial intelligence workloads, and observe that artificial intelligence workloads can cause performance degradation by generating a lot of I/Os (Inputs / Outputs) to NAND flash storage in mobile systems due to weak temporal locality and irregular popularity bias in memory write operations. Based on this observation, we discuss system architectures that can efficiently execute artificial intelligence workloads in mobile systems. Specifically, we adopt small persistent memory as write accelerator and show how efficiently the memory write operation of mobile systems can be managed. Simulation experiments show that the proposed system architecture can improve I/O time significantly compared to existing mobile systems.",

keywords = "AI workload, Mobile system, machine learning, memory reference, write acceleration",

author = "Seokmin Kwon and Hyokyung Bahn",

note = "Publisher Copyright: {\textcopyright} 2023 IEEE.; 2023 International Conference on Electrical and Information Technology, IEIT 2023 ; Conference date: 14-09-2023 Through 15-09-2023",

year = "2023",

doi = "10.1109/IEIT59852.2023.10335577",

language = "English",

series = "Proceedings - IEIT 2023: 2023 International Conference on Electrical and Information Technology",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

pages = "281--285",

booktitle = "Proceedings - IEIT 2023",

}

Kwon, S & Bahn, H 2023, Memory Reference Analysis and Implications for Executing AI Workloads in Mobile Systems. in Proceedings - IEIT 2023: 2023 International Conference on Electrical and Information Technology. Proceedings - IEIT 2023: 2023 International Conference on Electrical and Information Technology, Institute of Electrical and Electronics Engineers Inc., pp. 281-285, 2023 International Conference on Electrical and Information Technology, IEIT 2023, Malang, Indonesia, 14/09/23. https://doi.org/10.1109/IEIT59852.2023.10335577

Memory Reference Analysis and Implications for Executing AI Workloads in Mobile Systems. / Kwon, Seokmin; Bahn, Hyokyung.
Proceedings - IEIT 2023: 2023 International Conference on Electrical and Information Technology. Institute of Electrical and Electronics Engineers Inc., 2023. p. 281-285 (Proceedings - IEIT 2023: 2023 International Conference on Electrical and Information Technology).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - Memory Reference Analysis and Implications for Executing AI Workloads in Mobile Systems

AU - Kwon, Seokmin

AU - Bahn, Hyokyung

PY - 2023

Y1 - 2023

N2 - Recently, mobile applications that internally execute artificial intelligence workloads are increasing. In mobile environments, performance degradation due to memory thrashing may occur while artificial intelligence workloads perform the training of large data set due to limitations in memory capacity. In this paper, we analyze the memory reference characteristics of artificial intelligence workloads, and observe that artificial intelligence workloads can cause performance degradation by generating a lot of I/Os (Inputs / Outputs) to NAND flash storage in mobile systems due to weak temporal locality and irregular popularity bias in memory write operations. Based on this observation, we discuss system architectures that can efficiently execute artificial intelligence workloads in mobile systems. Specifically, we adopt small persistent memory as write accelerator and show how efficiently the memory write operation of mobile systems can be managed. Simulation experiments show that the proposed system architecture can improve I/O time significantly compared to existing mobile systems.

AB - Recently, mobile applications that internally execute artificial intelligence workloads are increasing. In mobile environments, performance degradation due to memory thrashing may occur while artificial intelligence workloads perform the training of large data set due to limitations in memory capacity. In this paper, we analyze the memory reference characteristics of artificial intelligence workloads, and observe that artificial intelligence workloads can cause performance degradation by generating a lot of I/Os (Inputs / Outputs) to NAND flash storage in mobile systems due to weak temporal locality and irregular popularity bias in memory write operations. Based on this observation, we discuss system architectures that can efficiently execute artificial intelligence workloads in mobile systems. Specifically, we adopt small persistent memory as write accelerator and show how efficiently the memory write operation of mobile systems can be managed. Simulation experiments show that the proposed system architecture can improve I/O time significantly compared to existing mobile systems.

KW - AI workload

KW - Mobile system

KW - machine learning

KW - memory reference

KW - write acceleration

UR - http://www.scopus.com/inward/record.url?scp=85181088087&partnerID=8YFLogxK

U2 - 10.1109/IEIT59852.2023.10335577

DO - 10.1109/IEIT59852.2023.10335577

M3 - Conference contribution

AN - SCOPUS:85181088087

T3 - Proceedings - IEIT 2023: 2023 International Conference on Electrical and Information Technology

SP - 281

EP - 285

BT - Proceedings - IEIT 2023

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 2023 International Conference on Electrical and Information Technology, IEIT 2023

Y2 - 14 September 2023 through 15 September 2023

ER -

Kwon S, Bahn H. Memory Reference Analysis and Implications for Executing AI Workloads in Mobile Systems. In Proceedings - IEIT 2023: 2023 International Conference on Electrical and Information Technology. Institute of Electrical and Electronics Engineers Inc. 2023. p. 281-285. (Proceedings - IEIT 2023: 2023 International Conference on Electrical and Information Technology). doi: 10.1109/IEIT59852.2023.10335577

Memory Reference Analysis and Implications for Executing AI Workloads in Mobile Systems

Abstract

Publication series

Conference

Bibliographical note

Keywords

Access to Document

Fingerprint

Cite this