深圳大学计算机与软件学院
College of Computer Science and Software Engineering, SZU

EBI-PAI: Towards An Efficient Edge-Based IoT Platform for

Artificial Intelligence

IEEE Internet of Things Journal 

 

Shu Yang    Kunkun Xu    Laizhong Cui    Zhongxing Ming    Ziteng Chen    Zhong Ming

Shenzhen University

 

Abstract

Edge computing, especially multiaccess edge computing, is seen as a promising technology to improve the Quality of user Experience (QoE) of many artificial intelligence (AI) applications in the evolution toward Internet-of-Things (IoT) infrastructure. However, the management and deployment of massive edge data centers bring new challenges for the current network. In this article, we propose a new edge-based IoT platform for AI (EBI-PAI), based on software-defined network (SDN) and serverless technology. EBI-PAI provides a unified service calling interface and schedules the resources automatically to satisfy the QoE requirements of users. To optimize performances during incremental deployment, we formulate the deployment problem, prove its complexity, and design heuristic algorithms to solve it. We implement EBI-PAI based on an opensource serverless project and deploy it in real networks. To evaluate EBI-PAI, we conduct comprehensive simulations based on the generated and real-world network topology, and real-world base station data set. The simulation results show that EBI-PAI can greatly improve QoE with the same budget and save the budget to achieve similar QoE. We finally carry out a case study with real user demands, and it further validates the simulation results.

Fig. 1. The difference between our controller and SDN controller is that the SDN controller customizes the forwarding table of routers and switches to control the forwarding path, while the EBI-PAI controller manipulates the path in the application layer.

 

Fig. 2 shows the edge computing platform deployed on an MEC server based on the serverless technology (such as OpenWhisk). The entry point of the platform is a trigger associated with a specific event.

 

Fig. 4. The controller is responsible for code distribution, status information collection, and user access control, and is composed of three modules: 1) access management; 2) distributor; and 3) platform gateway.

 

Fig. 17 shows the processing results of the Shanghai Telecom BS data set using the K-medoids clustering algorithm. The experiment sets the number of clusters to 20, that is, all BSs are divided into 20 clusters according to distance. A server is deployed on the central BS of each cluster so that the sum of the distance between the BS and all other BSs in the cluster is the smallest, The maximum distance from the cluster center to other nodes in the cluster is used as the QoE requirement that the algorithm can satisfy.

 

Fig. 18 shows the execution results of 100, 500, and 1000 concurrent requests of EBI-PAI and cloud-based solutions. We run each experiment ten times, calculate the average value

 

Acknowledgements

This work was supported in part by the National Key Research and Development Plan of China under Grant 2018YFB1800302 and Grant 2018YFB1800805; in part by the National Natural Science Foundation of China under Grant 61772345, Grant 61902258, Grant 61672358, and Grant 61836005; in part by the Major Fundamental Research Project in the Science and Technology Plan of Shenzhen under Grant JCYJ20190808142207420 and Grant GJHZ20190822095416463; and in part by the Pearl River Young Scholars Funding of Shenzhen University.

 

Bibtex

@ARTICLE{9174943,

author={Yang, Shu and Xu, Kunkun and Cui, Laizhong and Ming, Zhongxing and Chen, Ziteng and Ming, Zhong},

journal={IEEE Internet of Things Journal},

title={EBI-PAI: Toward an Efficient Edge-Based IoT Platform for Artificial Intelligence},

year={2021},

volume={8},

number={12},

pages={9580-9593},

doi={10.1109/JIOT.2020.3019008}

}

Downloads