Pipette: Efficient Fine-Grained Reads for SSDs

Shuhan Bai; Hu Wan; Yun Huang; Xuan Sun; Fei Wu; Changsheng Xie; Hung-Chih Hsieh; Tei-Wei Kuo; Chun Jason Xue

doi:10.1109/TCAD.2023.3276520

Pipette: Efficient Fine-Grained Reads for SSDs

Shuhan Bai
, Hu Wan
, Yun Huang
, Xuan Sun
, Fei Wu^*
, Changsheng Xie
, Hung-Chih Hsieh
, Tei-Wei Kuo
, Chun Jason Xue

^*Corresponding author for this work

Department of Computer Science

Research output: Journal Publications and Reviews › RGC 21 - Publication in refereed journal › peer-review

79 Downloads (CityUHK Scholars)

Abstract

Big data applications, such as recommendation system and social network, often generate a huge number of fine-grained reads to the storage. Block-oriented storage devices upon the traditional storage system rely on the paging mechanism to migrate pages to the host DRAM, tending to suffer from these fine-grained read operations in terms of I/O traffic as well as performance. Motivated by this challenge, an efficient fine-grained read framework, Pipette, is proposed in this paper as an extension to the traditional I/O framework. With adaptive design for caching, merging and scheduling, Pipette explores locality and acceleration for fine-grained read requests to establish an efficient byte-granular read path upon the dedicated byte-addressable interface. When the Pipette prototype on an SSD runs popular workloads, we measured throughput gains by up to 50% and 54% with traffic reduction in the range of 41.3× and 56.5×. © 2023 IEEE.

Original language	English
Pages (from-to)	4721-4734
Journal	IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems
Volume	42
Issue number	12
Online published	15 May 2023
DOIs	https://doi.org/10.1109/TCAD.2023.3276520
Publication status	Published - Dec 2023

Funding

This work was supported in part by the National Natural Science Foundation of China under Grant 61821003, Grant U2001203, Grant 61872413, and Grant 61902137; and in part by the Research Grants Council of the Hong Kong Special Administrative Region, China, under Grant CityU 11217020 and Grant CityU 11218720.

Research Keywords

File system
fine-grained reads
solid-state drive

Publisher's Copyright Statement

COPYRIGHT TERMS OF DEPOSITED POSTPRINT FILE: © 2023 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works. Bai, S., Wan, H., Huang, Y., Sun, X., Wu, F., Xie, C., Hsieh, H-C., Kuo, T-W., & Xue, C. J. (2023). Pipette: Efficient Fine-Grained Reads for SSDs. IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems, 42(12), 4721 - 4734. https://doi.org/10.1109/TCAD.2023.3276520

RGC Funding Information

RGC-funded

Access Output

10.1109/TCAD.2023.3276520

151372011.pdfPost-print, 4.33 MB

Other Access Options

Check CityUHK Library

Permanent Link

Right click to copy link

Cite this

@article{c9fd7dd112b940dbb74b5e303bf169bd,

title = "Pipette: Efficient Fine-Grained Reads for SSDs",

abstract = "Big data applications, such as recommendation system and social network, often generate a huge number of fine-grained reads to the storage. Block-oriented storage devices upon the traditional storage system rely on the paging mechanism to migrate pages to the host DRAM, tending to suffer from these fine-grained read operations in terms of I/O traffic as well as performance. Motivated by this challenge, an efficient fine-grained read framework, Pipette, is proposed in this paper as an extension to the traditional I/O framework. With adaptive design for caching, merging and scheduling, Pipette explores locality and acceleration for fine-grained read requests to establish an efficient byte-granular read path upon the dedicated byte-addressable interface. When the Pipette prototype on an SSD runs popular workloads, we measured throughput gains by up to 50\% and 54\% with traffic reduction in the range of 41.3× and 56.5×. {\textcopyright} 2023 IEEE. ",

keywords = "File system, fine-grained reads, solid-state drive",

author = "Shuhan Bai and Hu Wan and Yun Huang and Xuan Sun and Fei Wu and Changsheng Xie and Hung-Chih Hsieh and Tei-Wei Kuo and Xue, \{Chun Jason\}",

year = "2023",

month = dec,

doi = "10.1109/TCAD.2023.3276520",

language = "English",

volume = "42",

pages = "4721--4734",

journal = "IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems",

issn = "0278-0070",

publisher = "IEEE",

number = "12",

}

TY - JOUR

T1 - Pipette

T2 - Efficient Fine-Grained Reads for SSDs

AU - Bai, Shuhan

AU - Wan, Hu

AU - Huang, Yun

AU - Sun, Xuan

AU - Wu, Fei

AU - Xie, Changsheng

AU - Hsieh, Hung-Chih

AU - Kuo, Tei-Wei

AU - Xue, Chun Jason

PY - 2023/12

Y1 - 2023/12

N2 - Big data applications, such as recommendation system and social network, often generate a huge number of fine-grained reads to the storage. Block-oriented storage devices upon the traditional storage system rely on the paging mechanism to migrate pages to the host DRAM, tending to suffer from these fine-grained read operations in terms of I/O traffic as well as performance. Motivated by this challenge, an efficient fine-grained read framework, Pipette, is proposed in this paper as an extension to the traditional I/O framework. With adaptive design for caching, merging and scheduling, Pipette explores locality and acceleration for fine-grained read requests to establish an efficient byte-granular read path upon the dedicated byte-addressable interface. When the Pipette prototype on an SSD runs popular workloads, we measured throughput gains by up to 50% and 54% with traffic reduction in the range of 41.3× and 56.5×. © 2023 IEEE.

AB - Big data applications, such as recommendation system and social network, often generate a huge number of fine-grained reads to the storage. Block-oriented storage devices upon the traditional storage system rely on the paging mechanism to migrate pages to the host DRAM, tending to suffer from these fine-grained read operations in terms of I/O traffic as well as performance. Motivated by this challenge, an efficient fine-grained read framework, Pipette, is proposed in this paper as an extension to the traditional I/O framework. With adaptive design for caching, merging and scheduling, Pipette explores locality and acceleration for fine-grained read requests to establish an efficient byte-granular read path upon the dedicated byte-addressable interface. When the Pipette prototype on an SSD runs popular workloads, we measured throughput gains by up to 50% and 54% with traffic reduction in the range of 41.3× and 56.5×. © 2023 IEEE.

KW - File system

KW - fine-grained reads

KW - solid-state drive

UR - https://www.scopus.com/pages/publications/85162904924

UR - https://www.scopus.com/record/pubmetrics.uri?eid=2-s2.0-85162904924&origin=recordpage

U2 - 10.1109/TCAD.2023.3276520

DO - 10.1109/TCAD.2023.3276520

M3 - RGC 21 - Publication in refereed journal

SN - 0278-0070

VL - 42

SP - 4721

EP - 4734

JO - IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems

JF - IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems

IS - 12

ER -

Pipette: Efficient Fine-Grained Reads for SSDs

Abstract

Funding

Research Keywords

Publisher's Copyright Statement

RGC Funding Information

Access Output

Other Access Options

Permanent Link

Other Files and Links

Fingerprint

GRF: Cutting the Tail Latency for Low-Cost High-Density Server SSDs with Reliability Considerations

GRF: How to Utilize a Huge Number of Flash Chips to Meet the Performance and Reliability Requirements via Self-Healing and Partitioning

Accelerating Large-Scale Recommender Systems

Cite this

Pipette: Efficient Fine-Grained Reads for SSDs

Abstract

Funding

Research Keywords

Publisher's Copyright Statement

RGC Funding Information

Access Output

Other Access Options

Permanent Link

Other Files and Links

Fingerprint

Projects

GRF: Cutting the Tail Latency for Low-Cost High-Density Server SSDs with Reliability Considerations

GRF: How to Utilize a Huge Number of Flash Chips to Meet the Performance and Reliability Requirements via Self-Healing and Partitioning

Student theses

Accelerating Large-Scale Recommender Systems

Cite this