Tiresias: A GPU Cluster Manager for Distributed Deep Learning

Authors: 

Juncheng Gu, Mosharaf Chowdhury, and Kang G. Shin, University of Michigan, Ann Arbor; Yibo Zhu, Microsoft and Bytedance; Myeongjae Jeon, Microsoft and UNIST; Junjie Qian, Microsoft; Hongqiang Liu, Alibaba; Chuanxiong Guo, Bytedance

Abstract: 

Deep learning (DL) training jobs bring some unique challenges to existing cluster managers, such as unpredictable training times, an all-or-nothing execution model, and inflexibility in GPU sharing. Our analysis of a large GPU cluster in production shows that existing big data schedulers cause long queueing delays and low overall performance.

We present Tiresias, a GPU cluster manager tailored for distributed DL training jobs, which efficiently schedules and places DL jobs to reduce their job completion times (JCTs). Given that a DL job’s execution time is often unpredictable, we propose two scheduling algorithms – Discretized Two-Dimensional Gittins index relies on partial information and Discretized Two-Dimensional LAS is information-agnostic – that aim to minimize the average JCT. Additionally, we describe when the consolidated placement constraint can be relaxed, and present a placement algorithm to leverage these observations without any user input. Experiments on the Michigan ConFlux cluster with 60 P100 GPUs and large-scale trace-driven simulations show that Tiresias improves the average JCT by up to 5.5× over an Apache YARN-based resource manager used in production. More importantly, Tiresias’s performance is comparable to that of solutions assuming perfect knowledge.

Open Access Media

USENIX is committed to Open Access to the research presented at our events. Papers and proceedings are freely available to everyone once the event begins. Any video, audio, and/or slides that are posted after the event are also free and open to everyone. Support USENIX and our commitment to Open Access.

BibTeX
@inproceedings {227623,
author = {Juncheng Gu and Mosharaf Chowdhury and Kang G. Shin and Yibo Zhu and Myeongjae Jeon and Junjie Qian and Hongqiang Liu and Chuanxiong Guo},
title = {Tiresias: A {GPU} Cluster Manager for Distributed Deep Learning},
booktitle = {16th {USENIX} Symposium on Networked Systems Design and Implementation ({NSDI} 19)},
year = {2019},
isbn = {978-1-931971-49-2},
address = {Boston, MA},
pages = {485--500},
url = {https://www.usenix.org/conference/nsdi19/presentation/gu},
publisher = {{USENIX} Association},
}