Topology-Oriented Virtual Network Embedding Approach for Data Centers

Abstract

Currently, data centers have become an attractive candidate for users that require IT resources in the form of virtual networks to run their applications. Optimal mapping of the virtual network on the top of the substrate network with resource constraint is called virtual network embedding (VNE) problem. Most of the VNE algorithms are general algorithms for random topology and not suitable for data centers due to particular topological characteristics. To solve the VNE problem in data centers, this paper develops a topology-oriented algorithm based on the Discrete Particle Swarm Optimization (DPSO). We first develop a maximum spanning algorithm to compute the ranking of virtual nodes based on, not only its bandwidth and degree, but also its connectivity in the entire virtual network. Then, the virtual networks are embedded onto the substrate network according to the connectivity ranking result by a DPSO-based algorithm, in which we also propose a topological heuristic information of substrate network and combine it into the particle search process for boosting convergence speed and revenue/cost ratio of substrate network. The evaluation results show that the proposed algorithm can improve the optimization performance of VNE by comparing with a few existing algorithms. INDEX TERMS Data center, virtual resource allocation, virtual network embedding, discrete particle swarm optimization.

Optimal Virtual Network Embeddings for Tree Topologies

“…Theorem 1) yields algorithms of practical significance only when the number of request nodes is small and in our computational study we restrict our attention to request graphs on less than 12 nodes. While this may be considered to be an unreasonably small number of nodes, many existing works on the VNEP [6,9] and its applications in data centers [30,31] consider requests of such size.…”

Towards security-Aware 5G slice embedding

“…Typical approaches apply a greedy algorithm that maps the VNFs requiring larger resources to the substrate nodes providing larger resources. Next, virtual links are mapped to the substrate paths using the shortest or k-shortest pathCao et al (2017);Zhang et al (2018);Yuan et al (2019). In the one-stage slice embedding approach, virtual links are mapped at the same time as VNFs.…”