Research On Key Technologies For Virtual Network Embedding With High Utility And Low Energy Cost

To solve the existing problems in the mobility, reliability and safety of the Internet, network virtualization has emerged to serve as the foundation of the future Internet by decoupling the role of service provider from the role of the infrastructure provider and allowing multiple heterogeneous virtual networks to coexist on a shared substrate network. As one of the key challenges of network virtualization environment, virtual network embedding problem has received significant attention in recent years both at home and abroad. Through deep analysis of the problem, however, there exists a lot of problems in the utility and energy aspects of the virtual network embedding in the prior studies. To solve these problems, this thesis carries on the thorough research and makes several attributions which are generalized as follows:·This thesis proposes a particle swarm optimization based algorithm to improve the utility of the substrate network. The utility of the substrate network refers to how to make the best utilization of the limited resources in the substrate network, and how to embed more virtual network requests and obtain higher revenues. This thesis studies this problem in two as-pects:the formulation of single-domain virtual network embedding prob-lem and optimization algorithm. In particular, first, this thesis establishes two models for virtual network embedding depending on that the sub-strate network supports path splitting or not. Then, by optimizing the resource overhead of the substrate network, a unified meta heuristic based algorithm is proposed to improve the virtual network acceptance ratio and generate more revenues. ·This thesis proposes two energy-aware algorithms in performing single-domain virtual network embedding to minimize the energy consumption while maintaining high revenues for the substrate network. In the con-text of the single-domain virtual network embedding, the primary goal of prior work is to maximize the revenues by accommodating more vir-tual network requests. The key limitation of prior studies is that they did not consider the energy cost for serving virtual network requests, which significantly affects the economic benefits of the infrastructure provider. However, according to the recent related research, energy related cost has made up to40-50%of the operating costs of the infrastructure provider. In order to optimize the energy consumption of the substrate network, con-sidering the single-domain virtual network embedding context, this thesis first abstracts the node energy consumption model and link energy con-sumption model of the substrate network to accurately describe these two kinds of energy consumptions. Then, incorporating the goals of reducing energy consumption and improving the acceptance ratio of virtual net-works, this thesis applies the technique of virtual network consolidation and designs energy-aware virtual network embedding algorithms to min-imize the energy consumption while maintaining high revenues for the substrate network.·This thesis considers the energy cost and proposes energy cost-aware al-gorithms in performing multi-domain virtual network embedding. In the context of the multi-domain virtual network embedding, by taking multi-ple factors, such as utility, electricity and energy consumption, into con-sideration, this thesis first exploits the location-varying and time-varying diversities of the electricity price, then extends the energy consumption model of single-domain context into the corresponding model of multi-domain context, and finally puts forward electricity cost aware multi-domain virtual network embedding algorithms, to further optimize the electricity cost of the infrastructure provider.