Article Details

Abstract

Cloud-edge collaborative computing networks face challenges in guaranteeing service latency due to dynamic network changes, heterogeneous resources, and load fluctuations. Addressing the issues of poor adaptability and low resource utilization in existing static scheduling mechanisms, this paper proposes a low-latency guarantee mechanism integrating prediction and adaptive migration. This mechanism constructs a latency-aware resource model, employs deep reinforcement learning to implement dynamic offloading decisions, predicts load using LSTM, reserves buffer resources, and designs a multi-attribute scoring migration strategy to handle burst scenarios. Experiments demonstrate that the proposed mechanism reduces end-to-end latency by 19%～47% under different loads, keeps the key task violation rate within 1.2% and drops it to 0.3% in burst scenarios, stabilizes the resource utilization rate of edge nodes at 72%～78%, and keeps performance degradation to no more than 12% under network fluctuations, effectively enhancing service quality guarantee capability and system robustness.

Keywords

cloud-edge collaboration; computing power network; low-latency guarantee; task scheduling; reinforcement learning; service quality

References

1. [1] Song Yaqin, Xu Hui, Liu Xianfeng, etc Cloud edge collaborative network architecture and key technologies for 6G satellite ground fusion [J]. Integrated Information Network of Heaven and Earth, 2023, 4 (3): 3-11
2. [2] Chen Xingyan, Zhang Xuesong, Xie Zhilong, etc Joint optimization method for computing and transmission of "cloud edge end" computing power system [J]. Computer Research and Development, 2023, 60 (4): 719-734
3. [3] Wang Shanshan, Luo Zhiyong Research on Efficient Edge Offloading Based on Cloud Edge Collaborative Computing Power Scheduling [J]. Radio Communication Technology, 2023, 49 (1): 56-62
4. [4] Zhong Orange Analysis of Computing Network Deployment Solutions [J]. Data Communication, 2022 (6): 16-20
5. [5] Ye Qindan, Fan Guisheng, Huang Hengyang Exploration of Integrated Support Scheme and Application Scenarios for Computing Power Networks [J]. Frontiers of Data and Computing Development, 2022, 4 (6): 55-66
6. [6] Wang Shuling, Sun Jie, Wang Peng, etc Resource scheduling optimization in cloud edge collaboration [J]. Telecommunications Science, 2023, 39 (2): 163-170
7. [7] Jia Qingmin, Hu Yujiao, Zhang Huayu, etc Research on Deterministic Computing Power Networks [J]. Journal of Communications, 2022, 43 (10): 55-64
8. [8] Lu Juhui, Wu Wen, Lin Zhixian A review of research on cloud edge collaboration and task scheduling in China from the perspective of bibliometrics [J]. Frontiers of Data and Computing Development, 2022, 4 (4): 132-141
9. [9] Wang Lili Analysis of Computing Network Deployment Solutions [J]. Telecommunications Science, 2022, 38 (6): 172-180
10. [10] Tang Jing, Jie Yunpeng, Leibo, etc Discussion on key technologies of edge computing in the industrial Internet scenario [J]. Automation Expo, 2022, 39 (2): 34-37

About the authors

Mo Jialiang (1988-), male, Han ethnicity, from Enping, Guangdong, is a senior design project manager, engineer, and holds a Master's degree in Engineering. His research interests include artificial intelligence, cloud computing, and IT support systems.