Mahboobeh Zangiabady, Alberto Garcia-Robledo, Christian Aguilar-Fuster & Javier Rubio-Loyola

Abstract

Network virtualization is a promising technology for overcoming Internet ossification by enabling multiple Virtual Networks (VNs) to coexist on a shared substrate network. One critical aspect in NV environments is the capability of operators to allocate resources in the substrate network to support VNs in an optimal manner. This is known as Virtual Network Embedding (VNE). In the same context, online VN migration is the process meant to re-allocate components of a VN in real-time and seamlessly to the end-users. Although progress has been made to address VN migration, there has been little investigation on integral migration approaches assessed under different VN environment conditions. The main contribution of this paper is a VN migration framework that addresses the online VN migration problem holistically, namely considering different aspects that affect the efficiency of resource (re)allocations and the VNE acceptance ratios, such as migration policies, trigger conditions, and the CPU capacity requirements for Intermediate Substrate Nodes. An evaluation methodology is developed for analyzing the performance of the proposed framework on substrate infrastructures of different sizes and densities. Extensive software simulations on substrate networks of varying size (50 to 250 nodes) and link density (0.06 to 0.6) discover the migration-oriented parameters that contribute to enhance VNE up to 18.7%. We also compare the framework performance against two state-of-the-art mechanisms that improve online VNE while looking for VNE solutions and observed acceptance ratio enhancements up to 3× higher when using our framework on a physical network with 100 nodes and a density of 0.06.

Keywords

VNE; VN Migration; Multi-constrained Optimal Path Search; Virtual Network Functions