摘要：With the rapid development of 5G technology, the service demand in various application scenarios is continuously increasing. Mobile edge computing (MEC) has become a popular computing paradigm by placing services and corresponding computing resources to edge servers to satisfy the low latency demands of users. However, edge servers lack a stable infrastructure for protection and limited storage space and computing power. Considering the reliability and stability of the edge system, efficiently placing resources and offloading tasks to the edge servers has become an urgent challenge. In this paper, we consider resource placement and task offloading strategies under different time scales to optimize the service response time in a dynamic edge system environment. We established the Markov model to obtain a quantitative relationship between system reliability and latency, and analyze the time required for resource and task offloading. Then, we propose the resource placement and task offloading (RPTO) algorithms under different time scales based on deep reinforcement learning (DRL) techniques with the aim of minimizing the cost of service providers in the long term. The experimental results demonstrate that our approach effectively tackles the challenges of joint resource placement and task offloading in the MEC. ? 2024 IEEE.

ISSN号：2836-3876

卷、期、页：Proceedings of the IEEE International Conference on Web Services, ICWS

发表日期：2024-07-07

期刊分区（SCI为中科院分区）：无

收录情况：EI（工程索引）

发表期刊名称：Proceedings of the IEEE International Conference on Web Services, ICWS

参与作者：陈莹,Truong, Hong-Linh

通讯作者：梁晶语,冯子涵,高涵

第一作者：黄霁崴

论文类型：会议论文

论文概要：梁晶语,冯子涵,高涵,陈莹,黄霁崴,Truong, Hong-Linh，Deep Reinforcement Learning based Reliability-aware Resource Placement and Task Offloading in Edge Computing，Proceedings of the IEEE International Conference on Web Services, ICWS，2024，Proceedings of the IEEE International Conference on Web Services, ICWS

论文题目：Deep Reinforcement Learning based Reliability-aware Resource Placement and Task Offloading in Edge Computing