.png){kind=icon}
.png){kind=icon}
Return
A Dependability-Aware Elastic Resource Management Framework for Edge Cloud-Native Computing
.png){kind=icon}
PRE
AI
DOI：10.1016/j.jss.2026.112920.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
The dependability of services in edge cloud-native environments is critically challenged by dynamic workloads, resource heterogeneity, and the need for low-latency responses. Ensuring reliable performance under these conditions requires resilient and adaptive resource management strategies. To tackle this issue, the emerging edge cloud-native paradigm offers a potential solution. The ability to detect changes in the system state and dynamically orchestrate and deploy container resources in multi-application scenarios at the edge and the cloud is of significant academic and practical importance. This paper proposes an approach to multiple-scenario elastic resource management based on an environment simulator in the edge cloud-native environment, leveraging a scenario characterization simulator. To validate our approach, we perform large-scale experiments using various algorithms for elastic resource management, workload migration, and resource utilization. The results demonstrate that our approach thereby significantly enhancing system dependability by improving throughput, reducing response time violations (a key dependability metric), and optimizing resource utilization under fluctuating loads. In the study, we proposed an offline reinforcement learning method with the ability to explore and optimize unified resource scheduling strategies. By leveraging reinforcement learning algorithms that maximize long-term rewards through offline training, we transform the deployment strategies of different scenarios from independent to cooperative and mutually beneficial. The results can be used as a pattern for similar scenarios and applications.
Keywords:
Edge Cloud-Native
Elastic Resource Management
System Dependability
Reinforcement Learning
Container Orchestration
Journal
IF:
4.1
Papers:
5.4K
Citations:
8.4K