发现信息
- 作者:S. Manikandan, R. Nadhiya Banu
- 年份:2026
- 发表源:Scientific Reports
- OpenAlex ID:https://openalex.org/W7225617783
- DOI:https://doi.org/10.1038/s41598-026-78699-5
- arXiv:待核验
- 检索主题:recursive-improvement
- OpenAlex 引用数快照:0(2026-11-12)
摘要线索
Intelligent resource orchestration in large-scale distributed computing systems can be facilitated via Software-Defined Networking (SDN)-enabled cross-cloud infrastructures, which allow programmability, elasticity, and service-centric administration. OpenFlow prediction accuracy, resource utilization efficiency, and Quality of Service (QoS) stability are significantly reduced by network heterogeneity, dynamic traffic fluctuations, controller congestion, and sophisticated cyberattacks. Optimization and learning-based techniques mostly estimate traffic or balance loads without topology-aware intelligence, adaptive decision-making, or synchronized cyber-physical representations for robust cross-cloud orchestration. DTR-GAT-OPAL (Digital Twin and Graph Attention Reinforcement Learning Framework for Intelligent OpenFlow Prediction and Secure Cross-Cloud Resource Orchestration) addresses these existing limitations. The proposed framework includes a Digital Twin (DT) layer for real-time synchronization of cyber-physical states, a Graph Attention Network (GAT) for topology-aware learning of flow representations, and a Proximal Policy Optimization (PPO)-based reinforcement learning engine for adaptive controller and virtual machine orchestration. Likewise included are a federated trust-security module and an uncertainty-aware QoS reasoning method for robust decision support under dynamic workloads. The same multi-objective optimization strategy maximizes service reliability and orchestration robustness while minimizing control-plane instability, migration overhead, energy consumption, carbon intensity, and security vulnerability. DTR-GAT-OPAL outperforms state-of-the-art methods in heterogeneous CloudSim-SDN and Mininet environments in OpenFlow prediction accuracy (0.99%), controller response latency (31.4%), energy consumption (34.2%), migration overhead (32.8%), and security assurance (29.6%). These results prove DTR-GAT-OPAL can orchestrate scalable, robust, and intelligent cross-cloud SDN resources for next-generation distributed computing ecosystems.
核验与筛选
- 当前状态:待筛选、去重和一手来源核验。
- OpenAlex 元数据和引用数只用于发现,不作为正式引文或质量判断。
- 记录正式版本、venue、方法细节、实验协议和局限后,才迁入稳定论文卡片。