Smash: a compression benchmark with AI datasets from remote GPU virtualization systems

Cristian Peñaranda; Carlos Reaño; Federico Silla

doi:10.1007/978-3-031-15471-3_21

Smash: a compression benchmark with AI datasets from remote GPU virtualization systems

Cristian Peñaranda^*, Carlos Reaño, Federico Silla

^*Corresponding author for this work

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

2 Citations (Scopus)

Abstract

Remote GPU virtualization is a mechanism that allows GPU-accelerated applications to be executed in computers without GPUs. Instead, GPUs from remote computers are used. Applications are not aware of using a remote GPU. However, overall performance depends on the throughput of the underlying network connecting the application to the remote GPUs. One way to increase this bandwidth is to compress transmissions made within the remote GPU virtualization middleware between the application side and the GPU side.

In this paper we make a two-fold contribution. On the one hand, we present a new compression benchmark with more than 40 compression libraries. On the other hand, we have gathered the internal transmissions of a remote GPU virtualization middleware while executing 4 popular artificial intelligence applications. With these data we have created a new dataset for testing compression libraries in this specific domain. Both, the new compression benchmark and the new dataset are publicly available.

Original language	English
Title of host publication	Hybrid Artificial Intelligent Systems: 17th International Conference, HAIS 2022, proceedings
Editors	Pablo García Bringas, Hilde Pérez García, Francisco Javier Martínez de Pisón, José Ramón Villar Flecha, Alicia Troncoso Lora, Enrique A. de la Cal, Álvaro Herrero, Francisco Martínez Álvarez, Giuseppe Psaila, Héctor Quintián, Emilio Corchado
Publisher	Springer Cham
Pages	236-248
ISBN (Electronic)	9783031154713
ISBN (Print)	9783031154706
DOIs	https://doi.org/10.1007/978-3-031-15471-3_21
Publication status	Published - 12 Sept 2022
Externally published	Yes
Event	17th International Conference on Hybrid Artificial Intelligence Systems, HAIS 2022 - Salamanca, Spain Duration: 05 Sept 2022 → 07 Sept 2022

Publication series

Name	Lecture Notes in Computer Science: Lecture Notes in Artificial Intelligence (LNAI)
Volume	13469
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	17th International Conference on Hybrid Artificial Intelligence Systems, HAIS 2022
Country/Territory	Spain
City	Salamanca
Period	05/09/2022 → 07/09/2022

Keywords

Artificial Intelligence
Benchmark
Compression
GPU

ASJC Scopus subject areas

Theoretical Computer Science
General Computer Science

Access to Document

10.1007/978-3-031-15471-3_21Licence: Unspecified

Cite this

Peñaranda, C., Reaño, C., & Silla, F. (2022). Smash: a compression benchmark with AI datasets from remote GPU virtualization systems. In P. García Bringas, H. Pérez García, F. J. Martínez de Pisón, J. R. Villar Flecha, A. Troncoso Lora, E. A. de la Cal, Á. Herrero, F. Martínez Álvarez, G. Psaila, H. Quintián, & E. Corchado (Eds.), Hybrid Artificial Intelligent Systems: 17th International Conference, HAIS 2022, proceedings (pp. 236-248). (Lecture Notes in Computer Science: Lecture Notes in Artificial Intelligence (LNAI); Vol. 13469). Springer Cham. https://doi.org/10.1007/978-3-031-15471-3_21

Peñaranda, Cristian ; Reaño, Carlos ; Silla, Federico. / Smash: a compression benchmark with AI datasets from remote GPU virtualization systems. Hybrid Artificial Intelligent Systems: 17th International Conference, HAIS 2022, proceedings. editor / Pablo García Bringas ; Hilde Pérez García ; Francisco Javier Martínez de Pisón ; José Ramón Villar Flecha ; Alicia Troncoso Lora ; Enrique A. de la Cal ; Álvaro Herrero ; Francisco Martínez Álvarez ; Giuseppe Psaila ; Héctor Quintián ; Emilio Corchado. Springer Cham, 2022. pp. 236-248 (Lecture Notes in Computer Science: Lecture Notes in Artificial Intelligence (LNAI)).

@inproceedings{5afa28a69cfe43b285d645f62ca06f25,

title = "Smash: a compression benchmark with AI datasets from remote GPU virtualization systems",

abstract = "Remote GPU virtualization is a mechanism that allows GPU-accelerated applications to be executed in computers without GPUs. Instead, GPUs from remote computers are used. Applications are not aware of using a remote GPU. However, overall performance depends on the throughput of the underlying network connecting the application to the remote GPUs. One way to increase this bandwidth is to compress transmissions made within the remote GPU virtualization middleware between the application side and the GPU side. In this paper we make a two-fold contribution. On the one hand, we present a new compression benchmark with more than 40 compression libraries. On the other hand, we have gathered the internal transmissions of a remote GPU virtualization middleware while executing 4 popular artificial intelligence applications. With these data we have created a new dataset for testing compression libraries in this specific domain. Both, the new compression benchmark and the new dataset are publicly available.",

keywords = "Artificial Intelligence, Benchmark, Compression, GPU",

author = "Cristian Pe{\~n}aranda and Carlos Rea{\~n}o and Federico Silla",

year = "2022",

month = sep,

day = "12",

doi = "10.1007/978-3-031-15471-3_21",

language = "English",

isbn = "9783031154706",

series = "Lecture Notes in Computer Science: Lecture Notes in Artificial Intelligence (LNAI)",

publisher = "Springer Cham",

pages = "236--248",

editor = "{Garc{\'i}a Bringas}, Pablo and {P{\'e}rez Garc{\'i}a}, Hilde and {Mart{\'i}nez de Pis{\'o}n}, {Francisco Javier} and {Villar Flecha}, {Jos{\'e} Ram{\'o}n} and {Troncoso Lora}, Alicia and {de la Cal}, {Enrique A.} and {\'A}lvaro Herrero and {Mart{\'i}nez {\'A}lvarez}, Francisco and Giuseppe Psaila and H{\'e}ctor Quinti{\'a}n and Emilio Corchado",

booktitle = "Hybrid Artificial Intelligent Systems: 17th International Conference, HAIS 2022, proceedings",

note = "17th International Conference on Hybrid Artificial Intelligence Systems, HAIS 2022 ; Conference date: 05-09-2022 Through 07-09-2022",

}

Peñaranda, C, Reaño, C & Silla, F 2022, Smash: a compression benchmark with AI datasets from remote GPU virtualization systems. in P García Bringas, H Pérez García, FJ Martínez de Pisón, JR Villar Flecha, A Troncoso Lora, EA de la Cal, Á Herrero, F Martínez Álvarez, G Psaila, H Quintián & E Corchado (eds), Hybrid Artificial Intelligent Systems: 17th International Conference, HAIS 2022, proceedings. Lecture Notes in Computer Science: Lecture Notes in Artificial Intelligence (LNAI), vol. 13469, Springer Cham, pp. 236-248, 17th International Conference on Hybrid Artificial Intelligence Systems, HAIS 2022, Salamanca, Spain, 05/09/2022. https://doi.org/10.1007/978-3-031-15471-3_21

Smash: a compression benchmark with AI datasets from remote GPU virtualization systems. / Peñaranda, Cristian; Reaño, Carlos; Silla, Federico.
Hybrid Artificial Intelligent Systems: 17th International Conference, HAIS 2022, proceedings. ed. / Pablo García Bringas; Hilde Pérez García; Francisco Javier Martínez de Pisón; José Ramón Villar Flecha; Alicia Troncoso Lora; Enrique A. de la Cal; Álvaro Herrero; Francisco Martínez Álvarez; Giuseppe Psaila; Héctor Quintián; Emilio Corchado. Springer Cham, 2022. p. 236-248 (Lecture Notes in Computer Science: Lecture Notes in Artificial Intelligence (LNAI); Vol. 13469).

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

TY - GEN

T1 - Smash: a compression benchmark with AI datasets from remote GPU virtualization systems

AU - Peñaranda, Cristian

AU - Reaño, Carlos

AU - Silla, Federico

PY - 2022/9/12

Y1 - 2022/9/12

N2 - Remote GPU virtualization is a mechanism that allows GPU-accelerated applications to be executed in computers without GPUs. Instead, GPUs from remote computers are used. Applications are not aware of using a remote GPU. However, overall performance depends on the throughput of the underlying network connecting the application to the remote GPUs. One way to increase this bandwidth is to compress transmissions made within the remote GPU virtualization middleware between the application side and the GPU side. In this paper we make a two-fold contribution. On the one hand, we present a new compression benchmark with more than 40 compression libraries. On the other hand, we have gathered the internal transmissions of a remote GPU virtualization middleware while executing 4 popular artificial intelligence applications. With these data we have created a new dataset for testing compression libraries in this specific domain. Both, the new compression benchmark and the new dataset are publicly available.

AB - Remote GPU virtualization is a mechanism that allows GPU-accelerated applications to be executed in computers without GPUs. Instead, GPUs from remote computers are used. Applications are not aware of using a remote GPU. However, overall performance depends on the throughput of the underlying network connecting the application to the remote GPUs. One way to increase this bandwidth is to compress transmissions made within the remote GPU virtualization middleware between the application side and the GPU side. In this paper we make a two-fold contribution. On the one hand, we present a new compression benchmark with more than 40 compression libraries. On the other hand, we have gathered the internal transmissions of a remote GPU virtualization middleware while executing 4 popular artificial intelligence applications. With these data we have created a new dataset for testing compression libraries in this specific domain. Both, the new compression benchmark and the new dataset are publicly available.

KW - Artificial Intelligence

KW - Benchmark

KW - Compression

KW - GPU

U2 - 10.1007/978-3-031-15471-3_21

DO - 10.1007/978-3-031-15471-3_21

M3 - Conference contribution

AN - SCOPUS:85139022505

SN - 9783031154706

T3 - Lecture Notes in Computer Science: Lecture Notes in Artificial Intelligence (LNAI)

SP - 236

EP - 248

BT - Hybrid Artificial Intelligent Systems: 17th International Conference, HAIS 2022, proceedings

A2 - García Bringas, Pablo

A2 - Pérez García, Hilde

A2 - Martínez de Pisón, Francisco Javier

A2 - Villar Flecha, José Ramón

A2 - Troncoso Lora, Alicia

A2 - de la Cal, Enrique A.

A2 - Herrero, Álvaro

A2 - Martínez Álvarez, Francisco

A2 - Psaila, Giuseppe

A2 - Quintián, Héctor

A2 - Corchado, Emilio

PB - Springer Cham

T2 - 17th International Conference on Hybrid Artificial Intelligence Systems, HAIS 2022

Y2 - 5 September 2022 through 7 September 2022

ER -

Peñaranda C, Reaño C, Silla F. Smash: a compression benchmark with AI datasets from remote GPU virtualization systems. In García Bringas P, Pérez García H, Martínez de Pisón FJ, Villar Flecha JR, Troncoso Lora A, de la Cal EA, Herrero Á, Martínez Álvarez F, Psaila G, Quintián H, Corchado E, editors, Hybrid Artificial Intelligent Systems: 17th International Conference, HAIS 2022, proceedings. Springer Cham. 2022. p. 236-248. (Lecture Notes in Computer Science: Lecture Notes in Artificial Intelligence (LNAI)). doi: 10.1007/978-3-031-15471-3_21