The Research On Energy-Saving Solutions In Virtual Network

The development of information communication technology brings not only convenience to people’s life, but also great energy consumption and waste of network resources. Network virtualization technology, which enables multiple virtual resources to run in the shared substrate network, greatly improves the energy efficiency of substrate network. The key technology of network virtualization is virtual network embedding(VNE), which mainly deals with the problem of allocating substrate network resources to virtual network requests. Most of existing VNE algorithms aim to maximize the Internet service provider(ISP) revenue through the accommodation of virtual network requests as many as possible in the same substrate resources. As research is going deep and energy problem becomes increasingly serious, researchers start to consider energy problem in embedding algorithms and design virtual network energy saving solutions based on energy efficiency optimization.This paper firstly introduces the technical background of network virtualization and virtual network energy saving solutions. Then it classifies the energy saving technologies according to application and analyzes energy-saving methods, including sleeping mechanism, dynamic adaptation, scale computation, selectively connection and network ranking. Then on the basis of the methods discussed above, the paper proposes energy-aware VNE solution based on minimizing the number of activated network devices and proposes heuristic EA-VNE algorithm and approximately optimal meta-heuristic EA-VNE-EPSO algorithm. EA-VNE’s idea is to select substrate resources activated with low electricity price as the embedding objective and make unused nodes and links to sleep. EA-VNE-EPSO based on particle swarm optimization(PSO) algorithm regards VNE solution as a particle which updates itself through tracking the best point of the individual and the global. Finally an approximately optimal VNE solution is able to be found after iteration process. The experiment results show that the solution reflects significant advantages in the performance of long-term revenue, energy cost and average time. Then the paper studys real-time energy optimized VNE solution with the example of IP over WDM network which optimizes energy efficiency and considers optical layer in link embedding. The paper develops heuristic REOViNE algorithm which classifies nodes in VNRs, embeds nodes according to CPU capacity and adopts minimum hop algorithm to embed links. The experiment results show that the solution is doing better in the performance of energy cost and acceptance rate of VNRs when the number and load conform to some condition.