Study On Optimal Virtual Network Embedding Algorithms For Cost And Energy Efficiency

With the rapid development of applications such as cloud computing and relatedcommunication services, current Internet appears much more ossifications in its architecture. Inorder to eliminate the inadaptability, researchers have put forward network virtualization, whichallows heterogeneous virtual networks to share the same substrate network. Virtual networkembedding (VNE) is one of key technologies to network virtualization. Most of existing VNEalgorithms focus on the cost of embedding. As global nature environment and energy problemshave become increasingly serious, more attentions have been paid on energy efficient networking.This thesis focuses on the VNE optimization and related algorithm for both cost efficiency andenergy efficiency. Firstly, the thesis presents the state of art of VNE researching, and introducesnode splitting techniques into VNE. It proposes then a VNE problem by mixed integerprogramming (MIP) formulation aims to minimize the cost. A new VNE algorithm called VNE-NSis developed for solving this problem. Secondly, in the thesis, a mathametic expression of energyconsumption for proportional computing is designed, and a MIP model is developed for theenergy-objective VNE problem. A VNE algorithm called VNE-RA is given for solving this problem.Finally, a comparative experiment of VNE-RA, VNE-EA and VNE-Cost is discribed.Furthermore, this thesis takes NSFNET as an example to verify the effectiveness of theproposed algorithms by using LINGO. The experimental results show that VNE-NS can decreasethe cost by10.2percent and effectively increase the acceptance ratio. In the network adoptingproportional computing, VNE-RA is proved to be the most energy-savable algorithm. Comparingwith VNE-EA and VNE-Cost, VNE-RA reduces energy consumption by4percent and10.5percent.And, in the case with lightly loaded traffics, VNE-RA can not only save more energy, but also leadsto high acceptance ratio.