Abstract
As 5G technology matures, the transition from centralized cloud-native architectures to decentralized edge-native systems introduces significant networking challenges. This paper presents a novel framework that integrates Software-Defined Networking (SDN) with Kubernetes (K8s) to optimize routing in 5G cloud-edge environments. By employing a three-layer SDN framework and a specialized multi-objective optimization algorithm, TCCA, the proposed strategy achieves superior load balancing, reduced latency, and lower operational costs. Simulation results demonstrate that TCCA consistently outperforms traditional metaheuristic algorithms, providing a robust solution for real-time data processing at the edge.
Introduction
The shift toward "Edge Native" architectures in the 5G era is driven by the need for low-latency, high-bandwidth processing near the data source. However, the edge environment is characterized by hardware heterogeneity and geographic dispersion. Traditional networking often results in a "silo" effect between the physical underlay and the container overlay, leading to inefficient routing and complex management. This research addresses these issues by fusing SDN and K8s into a unified control plane that treats network and computing power as a single, orchestratable resource.
Proposed Architecture
The authors introduce a comprehensive cloud-edge collaboration framework consisting of several key components designed to bridge the gap between physical and container networks.
Three-Layer SDN Framework
- Southbound Adaptation Layer: Utilizes YANG models and an SDN coordinator to abstract differences between various hardware vendors (e.g., Cisco) and platforms (OpenStack, K8s).
- Network Service Choreographer: Manages complex tasks like VPC routing, VPNs, and traffic engineering, decoupling business logic from underlying hardware.
- Northbound API: Provides a standardized interface for applications to request network resources.
K8s and SDN Integration
The architecture integrates a microservice Service Mesh with the SDN controller. This allows for "multi-threshold detection," monitoring both network congestion and container health simultaneously. By decoupling context boundaries, the system achieves "location transparency," where services do not need to know the physical destination of packets; the SDN/K8s fusion handles the mapping automatically.
The TCCA Optimization Algorithm
The core of the proposed policy is the TCCA algorithm, a multi-objective optimization tool. It evaluates routing paths based on three primary factors: load balancing, total delay, and cost. The algorithm's performance is validated using metrics such as Inverted Generational Distance (IGD) and Hypervolume (HV).
Evaluation and Results
The paper evaluates the TCCA algorithm within a simulated 5G cloud-edge environment, benchmarking it against ten other well-known optimization algorithms, including NSGA2, the Whale Optimization Algorithm (WA), and the Artificial Bee Colony Algorithm (ABCA).
Key findings include:
- Load Balancing: TCCA maintained a stable distribution of traffic across nodes even as task volume increased to 1100 units.
- Cost Efficiency: The algorithm achieved a lower total cost curve, indicating more efficient resource utilization.
- Latency Reduction: TCCA resulted in the lowest total delay, which is critical for the real-time processing requirements of 5G edge applications.
Conclusion and Future Work
The integration of SDN and Kubernetes, supported by the TCCA algorithm, offers a promising path for optimizing 5G edge networks. By advocating for a Spine+Leaf physical architecture combined with VXLAN and Regional Interconnection (RI) controllers, the authors solve persistent issues like IP reuse and tenant isolation.
Future research will focus on validating these results in live, large-scale 5G carrier networks and investigating the computational overhead associated with the SDN coordinator and RI controller in resource-constrained environments.
Summary updated