Abstract
The Internet of Vehicles (IoV) demands massive, instantaneous data transfers to ensure safety and efficiency in smart cities. Managing these resources across multi-domain physical networks is a significant challenge. This paper introduces BA-VNE, a bandwidth-aware algorithm that combines multi-domain virtual network embedding with Particle Swarm Optimization (PSO). Experimental results demonstrate that BA-VNE significantly improves Virtual Network Request (VNR) acceptance rates and reduces latency by prioritizing high-capacity links during the embedding process.
The Challenge of Bandwidth in IoV
The Internet of Vehicles (IoV) serves as the backbone of modern smart cities, enabling autonomous vehicles to communicate with satellites, road sensors, and other vehicles in real-time. These applications are inherently "bandwidth-hungry"; a single autonomous vehicle requires massive data transfers to perceive its environment safely and make split-second decisions.
Traditional network architectures often struggle to schedule these resources efficiently, particularly when the physical network is fragmented across multiple administrative domains. This creates a complex multi-domain Virtual Network Embedding (VNE) problem: mapping virtual requests (such as a vehicle's data stream) onto a fragmented physical infrastructure while ensuring sufficient bandwidth to prevent delays or system failures.
The BA-VNE Algorithm
To address these challenges, the authors propose the Bandwidth Aware Multi-Domain Virtual Network Embedding (BA-VNE) algorithm. The approach is built on several key pillars:
- Centralized Multi-Domain Architecture: A global controller manages inter-domain links, while local controllers handle resources within specific physical domains. This structure balances the need for global optimization with local administrative privacy.
- Bandwidth-Based Candidate Selection: Unlike standard algorithms that prioritize node proximity, BA-VNE filters for substrate links that exceed the average available bandwidth. This ensures that only high-capacity paths are considered for embedding.
- Particle Swarm Optimization (PSO): To avoid local optima, the algorithm employs PSO—a heuristic mimicking the social behavior of bird flocking—to iteratively improve the mapping of virtual nodes to physical ones.
- Two-Stage Embedding: The process is divided into intra-domain mapping (within a single provider) and inter-domain mapping (connecting providers), prioritizing links with the highest remaining resources.
Experimental Evaluation
The study utilized simulated network topologies generated via the GT-ITM tool. The physical network consisted of four domains, each with 30 nodes and initial link bandwidths between 1000 and 3000 Mbps. Virtual requests were generated using a Poisson distribution to simulate the bursty traffic patterns typical of IoV scenarios.
The simulation results highlighted several key performance advantages for BA-VNE:
- Acceptance Rate: BA-VNE maintained a VNR acceptance rate of approximately 60%, whereas competing models like VNE-PSO and LID-VNE dropped to roughly 30% as node density increased.
- Latency: The average embedding delay for BA-VNE remained below 500ms, significantly outperforming the MC-VNM algorithm, which saw delays exceeding 850ms.
- Bandwidth Selection: As shown in Figure 2, the BA-VNE algorithm consistently selected links with higher average bandwidth than the physical domain average, validating its bandwidth-aware selection criteria.
Conclusion
BA-VNE offers a robust solution for the high-data demands of IoV by making bandwidth the primary filter in the resource-scheduling process. By combining this focus with the load-balancing capabilities of PSO, the algorithm achieves higher acceptance rates and lower latency. Future work will focus on addressing the communication overhead of centralized controllers and the impact of high vehicle mobility on virtual mapping stability.
Summary updated