Couper: DNN Model Slicing for Visual Analytics Containers at the Edge

被引:42
作者
Hsu, Ke-Jou [1 ]
Bhardwaj, Ketan [1 ]
Gavrilovska, Ada [1 ]
机构
[1] Georgia Inst Technol, Atlanta, GA 30332 USA
来源
SEC'19: PROCEEDINGS OF THE 4TH ACM/IEEE SYMPOSIUM ON EDGE COMPUTING | 2019年
关键词
edge computing; video analytics application; pipeline processing; computation offloading; deep neural network; DNN partitioning; container orchestration;
D O I
10.1145/3318216.3363309
中图分类号
TP [自动化技术、计算机技术];
学科分类号
0812 ;
摘要
Applications incorporating DNN-based visual analytics are growing in demand. This class of data-intensive and latency-sensitive workloads has an opportunity to benefit from the emerging edge computing tier. However, to decouple the growing resource demand of DNN models, from the characteristics and resource limitations of the infrastructure elements available at the edge, new methods are needed to quickly slice the DNNs into appropriately-sized components, and to deploy those DNN slices to be executed on the edge infrastructure stacks. This paper presents Couper, a practical solution that provides for quick creation of slices of production DNNs for visual analytics, and enables their deployment in contemporary container-based edge software stacks. Couper is evaluated with 7 production DNNs, under varying edge configurations.
引用
收藏
页码:179 / 194
页数:16
相关论文
共 54 条
[1]  
Al Sadah Mohamed, VOD QAT CHIEF OP OFF VOD QAT CHIEF OP OFF
[2]  
[Anonymous], 2018, USENIX WORKSH HOT TO
[3]  
[Anonymous], 2018, {production-grade container orchestration} - automated container deployment, scaling, and management
[4]  
[Anonymous], 2011, CISCO VISUAL NETWORK
[5]  
[Anonymous], 2018, DOCK BUILD SHIP RUN
[6]  
[Anonymous], 2018, Akraino Edge Stack
[7]  
AWS, 2018, AWS GREENGR AM WEB S AWS GREENGR AM WEB S
[8]  
Bhardwaj K., 2018, USENIX WORKSH HOT TO
[9]   Fast, scalable and secure onloading of edge functions using AirBox [J].
Bhardwaj, Ketan ;
Shih, Ming-Wei ;
Agarwal, Pragya ;
Gavrilovska, Ada ;
Kim, Taesoo ;
Schwan, Karsten .
2016 FIRST IEEE/ACM SYMPOSIUM ON EDGE COMPUTING (SEC 2016), 2016, :14-27
[10]  
Biookaghazadeh S., 2018, P USENIX WORKSH HOT, P1