arrow
返回

Optimizing Job Offloading Schedule for Collaborative DNN Inference

delete2024-04-01
delete6
PRE
AI
Y
Yubin Duan
X
Xiangyu Wu *
DOI:10.1109/TMC.2023.3276937delete
delete原文链接
delete原文求助
delete分享
delete收藏
摘要

摘要

En 中文
Deep Neural Networks (DNNs) have been widely deployed in mobile applications. DNN inference latency is a critical metric to measure the service quality of those applications. Collaborative inference is a promising approach for latency optimization, where partial inference workloads are offloaded from mobile devices to cloud servers. Model partition problems for collaborative inference have been well studied. However, little attention has been paid to optimizing offloading pipeline for multiple DNN inference jobs. In practice, mobile devices usually need to process multiple DNN inference jobs simultaneously. We propose to jointly optimize the DNN partitioning and pipeline scheduling for multiple inference jobs. We theoretically analyze the optimal scheduling conditions for homogeneous chain-structure DNNs. Based on the analysis, we proposed near-optimal partitioning and scheduling methods for chain-structure DNNs. We also extend those methods for general-structure DNNs. In addition, we extend our problem scenario to handle heterogeneous DNN inference jobs. A layer-level scheduling algorithm is proposed. Theoretical analyses show that our proposed method is optimal when computation graphs are tree-structure. Our joint optimization methods are evaluated in a real-world testbed. Experiment results show that our methods can significantly reduce the overall inference latency of multiple inference jobs compared to partition-only or schedule-only approaches.
Keyword:
Pipelines
Processor scheduling
Mobile handsets
Task analysis
Solid modeling
Computational modeling
Cloud computing
Collaborative DNN inference
job offloading
makespan minimization
mobile cloud computing
pipeline scheduling

期刊

IEEE Transactions on Mobile Computing 封面图
IEEE Transactions on Mobile Computing
IF:
9.2
论文数:
5.8K
被引数:
1.8W

机构

P
pennsylvania commonwealth system of higher education (pcshe)
学者数:
12.9W
论文数: 11.7W
被引数: 177
引用论文

引用论文

Abstract 13806: Catheter Ablation Improves Quasar-Arterial Spin Labeling Based Regional Brain Perfusion in the Posterior Cingulate Cortex and Auditory-verbal Memory Function in Elderly Patients With Atrial Fibrillation
err2020-11-17
err0
PREAI
errYasuko Tatewaki; Tatsushi Mutoh; Hirokazu Sato; Ryo Enoki; Akiko Kobayashi; Michiho Muranaka; Shuzo Yamamoto; Tetsuo Yagi; Yasuyuki Taki
err分享
err收藏
Chilling Tolerance in Arabidopsis Involves ALA1, a Member of a New Family of Putative Aminophospholipid Translocases
err2000-12-01
err0
PREAI
errEric Gomes; Mia Kyed Jakobsen; Kristian B. Axelsen; Markus Geisler; Michael Gjedde Palmgren
err分享
err收藏
Performance Study of Screen-Printed Textile Antennas after Repeated Washing丝网印刷纺织品天线反复洗涤后的性能研究
err2014-06-17
err0
errOAAI
errI. Kazani; F. Declercq; M. L. Scarpello; C. Hertleer; H. Rogier; D. Vande Ginste; G. De Mey; G. Guxho; L. Van Langenhove
err分享
err收藏
Comparative evaluation of grip and pinch strength in an Asian and European population
err2012-02-02
err0
PREAI
errSyed Neshat Anjum; Poorva Choudary; Ravi Dimri; Sudhi Ankarath
err分享
err收藏
err分享
err收藏
学者 查看更多内容