Fine-Grained Virtual Network Embedding in 5G Hierarchical Network Infrastructure

Ling-Shu Li, Caixia Liu, Quan Yuan, Xiaoyang Fang, Yangyang Zhang2017 · DOI

Abstract

Resource allocation and deployment of network Service Function Chain (SFC) is a vital problem of 5G network virtualization. By introducing SDN and NFV technology, the core network is reclassified as the control layer, forwarding layer and scheduling layer. A Virtual Network Embedding algorithm based on subgraph isomorphism (VNE-MSI) is proposed to allocate resource of each layer in isolated way. The proposed heuristic algorithm collaboratively maps decoupling Service Function Chain with backtracking strategy. Collaborative search strategy is designed according to the characters of virtual Evolved Packet Core. The experimental evaluations show that this method especially performs well in acceptance ratio and revenue-to-cost ratio.

Introduction

As the telecommunications industry transitions toward 5G, traditional rigid network architectures have become a significant bottleneck. To achieve the required flexibility and scalability, 5G leverages Network Function Virtualization (NFV) and Software Defined Networking (SDN) to create Virtual Mobile Networks (VMNs). A core challenge in this environment is Virtual Network Embedding (VNE), which involves mapping virtual nodes and links onto a physical substrate of data centers and connections. This problem is NP-hard, and traditional two-stage algorithms—which map nodes and links sequentially—often suffer from "link depletion," where nodes are mapped but no viable bandwidth remains to connect them.

Hierarchical 5G Architecture

To address these challenges, this research proposes a hierarchical 5G core network architecture. By decoupling the Evolved Packet Core (EPC) into three distinct layers—a logic control layer, a forwarding (resource) layer, and a scheduling layer—the system gains fine-grained control over resource allocation.

The proposed 5G hierarchical network infrastructure including MANO and distributed data centers.

This decoupling allows for greater elasticity. For instance, if data traffic spikes, the orchestrator can scale or remap the forwarding layer independently of the control logic, ensuring service continuity without reconfiguring the entire service chain.

The VNE-MSI Algorithm

The proposed Virtual Network Embedding algorithm based on Modified Subgraph Isomorphism (VNE-MSI) utilizes a "one-phase" mapping approach. Unlike traditional methods, VNE-MSI maps nodes and links simultaneously, significantly reducing the risk of link-mapping failures.

The algorithm incorporates three key strategies:

  1. Modified Subgraph Isomorphism: By exploiting the inherent logical flow of Service Function Chains (SFCs), the algorithm uses a modified pruning strategy to narrow the search space, making the computationally intensive subgraph isomorphism problem manageable for real-time applications.
  2. Backtracking Strategy: This allows the algorithm to recover from local optima or dead ends during the mapping process.
  3. MAX-MIN Resource Rule: When selecting physical nodes, the algorithm prioritizes those whose available resources most closely match the request. This "tight fit" approach minimizes resource fragmentation, leaving larger contiguous blocks of resources available for future, more demanding requests.
Mapping of decoupled Service Function Chains (SFCs) for control and forwarding layers onto a physical substrate.

Experimental Evaluation

The VNE-MSI algorithm was validated using NS-3 simulations on a substrate network of 50 physical nodes. It was compared against established benchmarks: VF2A, SVNE, and Ullmann's algorithm.

Key findings include:

  • Acceptance Ratio: VNE-MSI maintained a consistently higher request acceptance ratio over time. As the physical network reached high utilization, VNE-MSI's holistic consideration of nodes and links allowed it to find valid embeddings that other algorithms missed.
  • Revenue-to-Cost Ratio: The algorithm achieved a superior revenue-to-cost ratio, peaking near 1.25. This indicates that Infrastructure Providers (InPs) can generate more revenue per unit of physical resource consumed compared to traditional mapping strategies.

Conclusion

The VNE-MSI algorithm provides a robust solution for the deployment of 5G Service Function Chains. By combining a hierarchical SDN-based architecture with a one-phase subgraph isomorphism approach, it achieves high resource efficiency and request acceptance. Future work will focus on refining pruning strategies for even larger, more dynamic edge computing environments.

Summary updated

Abstract

Resource allocation and deployment of network Service Function Chain (SFC) is a vital problem of 5G network virtualization. By introducing SDN and NFV technology, the core network is reclassified as the control layer, forwarding layer and scheduling layer. A Virtual Network Embedding algorithm based on subgraph isomorphism (VNE-MSI) is proposed to allocate resource of each layer in isolated way. The proposed heuristic algorithm collaboratively maps decoupling Service Function Chain with backtracking strategy. Collaborative search strategy is designed according to the characters of virtual Evolved Packet Core. The experimental evaluations show that this method especially performs well in acceptance ratio and revenue-to-cost ratio.
BibTeX
@misc{li2017fine,
  author = {Ling-Shu Li and Caixia Liu and Quan Yuan and Xiaoyang Fang and Yangyang Zhang},
  title = {{Fine-Grained Virtual Network Embedding in 5G Hierarchical Network Infrastructure}},
  year = {2017},
  doi = {10.2991/icwcsn-16.2017.66},
}

Loading related content...

Figure