A model for virtual network embedding across multiple infrastructure providers using genetic algorithm

Abstract

Network virtualization is an important aspect in cloud computing, where network is assumed to be consisting of infinite amount of nodes and links. The substrate network (physical network), has limited capacity and so virtual network embedding on the substrate network becomes a problem. Virtual network embedding is a computationally hard problem, considering various constraints on nodes and links. The proposed work applies Genetic Algorithm for the virtual network embedding problem for mapping multiple virtual network requests on infrastructure providers managing multiple substrate networks. Performance evaluation, through simulations, indicates that the proposed model preforms better for the performance metrics such as infrastructure provider revenue, acceptance ratio and node and link utilization in comparison to few other contemporary mapping models.