EC-MASS: Towards an efficient edge computing-based multi-video scheduling system

• Published in: Computer communications Vol. 193; pp. 355 - 364
• Main Authors: Yang, Shu, Dong, Qingzhen, Cui, Laizhong, Chen, Xun, Lei, Siyu, Wu, Yulei, Luo, Chengwen
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.09.2022
• Subjects: Edge computing; Task scheduling; Video analytics; Task scheduling; Video analytics; Edge computing
• ISSN: 0140-3664; 1873-703X
• DOI: 10.1016/j.comcom.2022.07.002

Video cameras have been deployed widely today. Although existing systems aim to optimize live video analytics from a variety of perspectives, they are agnostic to the workload dynamics in real-world. We propose EC-MASS, an edge computing-based video scheduling system achieving both cost and performance optimization with multiple cameras and edge data centers. The intuition behind EC-MASS is to adaptively map cameras to different edge data centers according to dynamically updated configurations of cameras. We prove that generating the optimal mapping scheduling scheme is NP-Complete, and develop the scheduling algorithm by leveraging the insights of the economy consideration of camera allocation. Using the algorithm, EC-MASS is able to balance the workload among edge data centers while reducing the cost of video analytics system. We evaluate EC-MASS with datasets of video configurations from real-world cameras which randomly generate configurations for cameras, with a testbed that consists of 60 cameras and 4 edge data centers. Our results show that EC-MASS consistently outperforms the status quo in terms of cost and performance stability.