南方科技大学知识苑(SUSTech KC): PPS: Fair and efficient black-box scheduling for multi-tenant GPU clusters

题名	PPS: Fair and efficient black-box scheduling for multi-tenant GPU clusters
作者	Ma，Kaihao 1; Cai，Zhenkun 1; Yan，Xiao2 ; Zhang，Yang 3; Liu，Zhi 1; Feng，Yihui 3; Li，Chao 3; Lin，Wei 3; Cheng，James 1
通讯作者	Yan，Xiao
发表日期	2024-06-01
DOI	10.1016/j.parco.2024.103082
发表期刊	Parallel Computing 影响因子和分区
ISSN	0167-8191
卷号	120
摘要	Multi-tenant GPU clusters are common, where users purchase GPU quota to run their neural network training jobs. However, strict quota-based scheduling often leads to cluster under-utilization, while allowing quota groups to use excess GPUs improves utilization but results in fairness problems. We propose PPS, a probabilistic prediction based scheduler, which uses job history statistics to predict future cluster status for making good scheduling decisions. Different from existing schedulers that rely on deep learning frameworks to adjust bad scheduling decisions and/or require detailed job information, PPS treats jobs as black boxes in that PPS runs a job to completion without adjustment once scheduled and requires only aggregate job statistics. The black-box feature is favorable due to its good generality, compatibility and security, and made possible by the predictability of aggregate resource utilization statistics of large clusters. Extensive experiments show that PPS achieves high cluster utilization and good fairness simultaneously.
关键词	Cloud computing Cluster scheduling GPU cluster Model training Multi-tenant Resource sharing
相关链接	[Scopus记录]
收录类别	SCI ; EI
语种	英语
学校署名	通讯
ESI学科分类	COMPUTER SCIENCE
Scopus记录号	2-s2.0-85189175736
来源库	Scopus
引用统计
成果类型	期刊论文
条目标识符	http://sustech.caswiz.com/handle/2SGJ60CL/741113
专题	工学院_计算机科学与工程系
作者单位	1.Department of Computer Science and Engineering,The Chinese University of Hong Kong,Hong Kong,Hong Kong 2.Department of Computer Science and Engineering,Southern University of Science and Technology,Shen Zhen,China 3.Alibaba Group,Beijing,China
通讯作者单位	计算机科学与工程系
推荐引用方式 GB/T 7714	Ma，Kaihao,Cai，Zhenkun,Yan，Xiao,et al. PPS: Fair and efficient black-box scheduling for multi-tenant GPU clusters[J]. Parallel Computing,2024,120.
APA	Ma，Kaihao.,Cai，Zhenkun.,Yan，Xiao.,Zhang，Yang.,Liu，Zhi.,...&Cheng，James.(2024).PPS: Fair and efficient black-box scheduling for multi-tenant GPU clusters.Parallel Computing,120.
MLA	Ma，Kaihao,et al."PPS: Fair and efficient black-box scheduling for multi-tenant GPU clusters".Parallel Computing 120(2024).