.png){kind=icon}
.png){kind=icon}
Return
Enhancing machine learning workloads using optimized cloud-edge computing architectures with service function chaining
.png){kind=icon}
PRE
AI
DOI：10.1016/j.swevo.2026.102479.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
The rapid growth of IoT applications produces massive data volumes that create challenges for traditional cloud-only processing due to high latency, bandwidth boundaries and privacy concerns. Cloud edge computing addresses these issues by moving computation closer to data sources; heterogeneous edge resources, dynamic workloads and limited device capabilities require intellectual organization. This study is motivated by the need for a scalable and efficient framework that can improve machine learning workloads across cloud and edge environments. The proposed framework integrates Dynamic Service Function Chaining (DYN-SFC), Hybrid GA-SJF scheduling, APEO-based presentation optimization, and SecScale blockchain security to achieve adaptive task achievement, effective resource application and secure declaration. The main objectives are to reduce potential, improve QoS, optimize workload distribution and enhance scalability. The framework completes enhanced routing efficiency, reduced bandwidth consumption and reliable AI performance, providing a bridge between resource-constrained edge devices and powerful cloud infrastructures for next-generation submissions. The proposed methodology integrates DYN-SFC, Hybrid GA-SJF scheduling, APEO-based adaptive optimization, and SecScale blockchain security to improve cloud edge computing performance. The framework dynamically optimizes VNF placement, workload allocation, resource presentation and secure communication across assorted nodes. Investigational findings show that the method realizes 42 ms latency, 958 tasks/s throughput, 95.4% QoS, 18.6 ms waiting time and stable scalability up to 500 nodes with <12% security overhead. The results confirm improved competence, dependability and flexibility compared with existing methods. Future work will explore reinforcement learning-based resource forecast, real-world edge personality, multi-cloud surroundings and large-scale IoT submissions for further scalability and energy optimization.
Journal
IF:
8.5
Papers:
2.1K
Citations:
1.0W