Characterizing Network Requirements for GPU API Remoting in AI Applications

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Wang, Tianxia, Chen, Zhuofu, Wei, Xingda, Gu, Jinyu, Chen, Rong, Chen, Haibo
Format: Preprint
Veröffentlicht: 2024
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866910306640330752
author Wang, Tianxia
Chen, Zhuofu
Wei, Xingda
Gu, Jinyu
Chen, Rong
Chen, Haibo
author_facet Wang, Tianxia
Chen, Zhuofu
Wei, Xingda
Gu, Jinyu
Chen, Rong
Chen, Haibo
contents GPU remoting is a promising technique for supporting AI applications. Networking plays a key role in enabling remoting. However, for efficient remoting, the network requirements in terms of latency and bandwidth are unknown. In this paper, we take a GPU-centric approach to derive the minimum latency and bandwidth requirements for GPU remoting, while ensuring no (or little) performance degradation for AI applications. Our study including theoretical model demonstrates that, with careful remoting design, unmodified AI applications can run on the remoting setup using commodity networking hardware without any overhead or even with better performance, with low network demands.
format Preprint
id arxiv_https___arxiv_org_abs_2401_13354
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Characterizing Network Requirements for GPU API Remoting in AI Applications
Wang, Tianxia
Chen, Zhuofu
Wei, Xingda
Gu, Jinyu
Chen, Rong
Chen, Haibo
Operating Systems
Networking and Internet Architecture
GPU remoting is a promising technique for supporting AI applications. Networking plays a key role in enabling remoting. However, for efficient remoting, the network requirements in terms of latency and bandwidth are unknown. In this paper, we take a GPU-centric approach to derive the minimum latency and bandwidth requirements for GPU remoting, while ensuring no (or little) performance degradation for AI applications. Our study including theoretical model demonstrates that, with careful remoting design, unmodified AI applications can run on the remoting setup using commodity networking hardware without any overhead or even with better performance, with low network demands.
title Characterizing Network Requirements for GPU API Remoting in AI Applications
topic Operating Systems
Networking and Internet Architecture
url https://arxiv.org/abs/2401.13354