Uncertain Virtual Network Embedding Using Neighborhood Information Based Edge Prediction

Network virtualization is a key technology for efficient resource sharing in modern data centers, particularly with the advent of paradigms like Software-Defined Networking (SDN) that enable flexible and centralized control. However, virtual network users cannot always express their exact requirements, leading to uncertainties in topology and resource demands. We model such requests as uncertain virtual networks (UVNs). In this paper, we describe such networks as uncertain virtual networks (UVN) and use uncertain graphs to model them. This paper proposes UVNE (uncertain virtual network embedding), which is a three-step algorithm to embed the UVNs. (۱) Extracting a certain virtual network from multiple versions of an uncertain virtual network using a proposed edge prediction algorithm based on the SOM (self-organizing map) classifier. (۲) Clustering the extracted virtual network with the HCS (highly connected subgraph) clustering algorithm. (۳) Embedding the extracted virtual network with a one-step embedding algorithm. The proposed algorithm is compared with edge prediction and virtual network embedding algorithms. The results demonstrate the strength of the edge prediction algorithm, the benefit, and the reduction of the cost of the embedding.

کلیدواژه ها:

Uncertain virtual network ، Edge prediction ، SOM classifier ، Neighborhood information ، Graph clustering