CUDA virtualization and remoting for GPGPU based acceleration offloading at the edge

Antonio Mentone, Diana Di Luccio, Luca Landolfi, Sokol Kosta, Raffaele Montella

Publikation: Bidrag til bog/antologi/rapport/konference proceedingKonferenceartikel i proceedingForskningpeer review

Resumé

In the last decade, GPGPU virtualization and remoting have been among the most hyped research topics in the field of computer science and engineering due to the rising of cloud computing based technologies. Public, private, and hybrid infrastructures need such virtualization tools in order to multiplex and better organize the computing resources.
With the advent of novel technologies and paradigms, such as edge computing, code offloading in mobile clouds, deep learning techniques, etc., the need for computing power, especially of specialized hardware such as GPUs, has skyrocketed.
Although many GPGPU virtualization tools are available nowadays, in this paper we concentrate on improving GVirtuS, our solution for GPU virtualization.
The contributions in this work focus on the CUDA plug-in, in order to provide updated performance enabling the next generation of GPGPU code offloading applications. Moreover, we present a new GVirtuS implementation characterized by a highly modular approach with a full multithread support. We evaluate and discuss the benchmarks of the new implementation comparing and contrasting the results versus the pure CUDA, the previous version of GVirtuS, and rCUDA.
OriginalsprogEngelsk
Titel The 12th International Conference on Internet and Distributed Computing Systems
Antal sider10
Vol/bind11874
ForlagSpringer
Publikationsdato10 nov. 2019
Sider 414-423
ISBN (Trykt)978-3-030-34914-1, 978-3-030-34913-4
ISBN (Elektronisk)978-3-030-34914-1
DOI
StatusUdgivet - 10 nov. 2019
BegivenhedThe 12th International Conference on Internet and Distributed Computing Systems - Naples, Italien
Varighed: 10 okt. 201912 okt. 2019
Konferencens nummer: 12
https://idcs2019.uniparthenope.it/

Konference

KonferenceThe 12th International Conference on Internet and Distributed Computing Systems
Nummer12
LandItalien
ByNaples
Periode10/10/201912/10/2019
Internetadresse
NavnLecture Notes in Computer Science
ISSN0302-9743

Fingerprint

Cloud computing
Computer science
Hardware
Virtualization
Graphics processing unit
Deep learning

Citer dette

Mentone, A., Di Luccio, D., Landolfi, L., Kosta, S., & Montella, R. (2019). CUDA virtualization and remoting for GPGPU based acceleration offloading at the edge. I The 12th International Conference on Internet and Distributed Computing Systems (Bind 11874, s. 414-423). Springer. Lecture Notes in Computer Science https://doi.org/10.1007/978-3-030-34914-1_39
Mentone, Antonio ; Di Luccio, Diana ; Landolfi, Luca ; Kosta, Sokol ; Montella, Raffaele. / CUDA virtualization and remoting for GPGPU based acceleration offloading at the edge. The 12th International Conference on Internet and Distributed Computing Systems . Bind 11874 Springer, 2019. s. 414-423 (Lecture Notes in Computer Science).
@inproceedings{88cd2b3beed24266b644f1433ff9d9f9,
title = "CUDA virtualization and remoting for GPGPU based acceleration offloading at the edge",
abstract = "In the last decade, GPGPU virtualization and remoting have been among the most hyped research topics in the field of computer science and engineering due to the rising of cloud computing based technologies. Public, private, and hybrid infrastructures need such virtualization tools in order to multiplex and better organize the computing resources.With the advent of novel technologies and paradigms, such as edge computing, code offloading in mobile clouds, deep learning techniques, etc., the need for computing power, especially of specialized hardware such as GPUs, has skyrocketed.Although many GPGPU virtualization tools are available nowadays, in this paper we concentrate on improving GVirtuS, our solution for GPU virtualization.The contributions in this work focus on the CUDA plug-in, in order to provide updated performance enabling the next generation of GPGPU code offloading applications. Moreover, we present a new GVirtuS implementation characterized by a highly modular approach with a full multithread support. We evaluate and discuss the benchmarks of the new implementation comparing and contrasting the results versus the pure CUDA, the previous version of GVirtuS, and rCUDA.",
author = "Antonio Mentone and {Di Luccio}, Diana and Luca Landolfi and Sokol Kosta and Raffaele Montella",
year = "2019",
month = "11",
day = "10",
doi = "10.1007/978-3-030-34914-1_39",
language = "English",
isbn = "978-3-030-34914-1",
volume = "11874",
series = "Lecture Notes in Computer Science",
publisher = "Springer",
pages = "414--423",
booktitle = "The 12th International Conference on Internet and Distributed Computing Systems",
address = "Germany",

}

Mentone, A, Di Luccio, D, Landolfi, L, Kosta, S & Montella, R 2019, CUDA virtualization and remoting for GPGPU based acceleration offloading at the edge. i The 12th International Conference on Internet and Distributed Computing Systems . bind 11874, Springer, Lecture Notes in Computer Science, s. 414-423, Naples, Italien, 10/10/2019. https://doi.org/10.1007/978-3-030-34914-1_39

CUDA virtualization and remoting for GPGPU based acceleration offloading at the edge. / Mentone, Antonio; Di Luccio, Diana; Landolfi, Luca; Kosta, Sokol; Montella, Raffaele.

The 12th International Conference on Internet and Distributed Computing Systems . Bind 11874 Springer, 2019. s. 414-423 (Lecture Notes in Computer Science).

Publikation: Bidrag til bog/antologi/rapport/konference proceedingKonferenceartikel i proceedingForskningpeer review

TY - GEN

T1 - CUDA virtualization and remoting for GPGPU based acceleration offloading at the edge

AU - Mentone, Antonio

AU - Di Luccio, Diana

AU - Landolfi, Luca

AU - Kosta, Sokol

AU - Montella, Raffaele

PY - 2019/11/10

Y1 - 2019/11/10

N2 - In the last decade, GPGPU virtualization and remoting have been among the most hyped research topics in the field of computer science and engineering due to the rising of cloud computing based technologies. Public, private, and hybrid infrastructures need such virtualization tools in order to multiplex and better organize the computing resources.With the advent of novel technologies and paradigms, such as edge computing, code offloading in mobile clouds, deep learning techniques, etc., the need for computing power, especially of specialized hardware such as GPUs, has skyrocketed.Although many GPGPU virtualization tools are available nowadays, in this paper we concentrate on improving GVirtuS, our solution for GPU virtualization.The contributions in this work focus on the CUDA plug-in, in order to provide updated performance enabling the next generation of GPGPU code offloading applications. Moreover, we present a new GVirtuS implementation characterized by a highly modular approach with a full multithread support. We evaluate and discuss the benchmarks of the new implementation comparing and contrasting the results versus the pure CUDA, the previous version of GVirtuS, and rCUDA.

AB - In the last decade, GPGPU virtualization and remoting have been among the most hyped research topics in the field of computer science and engineering due to the rising of cloud computing based technologies. Public, private, and hybrid infrastructures need such virtualization tools in order to multiplex and better organize the computing resources.With the advent of novel technologies and paradigms, such as edge computing, code offloading in mobile clouds, deep learning techniques, etc., the need for computing power, especially of specialized hardware such as GPUs, has skyrocketed.Although many GPGPU virtualization tools are available nowadays, in this paper we concentrate on improving GVirtuS, our solution for GPU virtualization.The contributions in this work focus on the CUDA plug-in, in order to provide updated performance enabling the next generation of GPGPU code offloading applications. Moreover, we present a new GVirtuS implementation characterized by a highly modular approach with a full multithread support. We evaluate and discuss the benchmarks of the new implementation comparing and contrasting the results versus the pure CUDA, the previous version of GVirtuS, and rCUDA.

U2 - 10.1007/978-3-030-34914-1_39

DO - 10.1007/978-3-030-34914-1_39

M3 - Article in proceeding

SN - 978-3-030-34914-1

SN - 978-3-030-34913-4

VL - 11874

T3 - Lecture Notes in Computer Science

SP - 414

EP - 423

BT - The 12th International Conference on Internet and Distributed Computing Systems

PB - Springer

ER -

Mentone A, Di Luccio D, Landolfi L, Kosta S, Montella R. CUDA virtualization and remoting for GPGPU based acceleration offloading at the edge. I The 12th International Conference on Internet and Distributed Computing Systems . Bind 11874. Springer. 2019. s. 414-423. (Lecture Notes in Computer Science). https://doi.org/10.1007/978-3-030-34914-1_39