Study On Resource Mapping Algorithms For Virtual Network Deployment

In order to overcome the ossification of the current Internet technology, network virtualization makes use of the technique of virtualization to share the same underlying physical network. Virtual network mapping/embedding (VNE) can meet a variety of user-level application requirements with diversity. VNE promoises to take advantage of the method of optimization techniques to deploy virtual networks to physical networks according to different target in application area. VNE issue contains a number of research hotspot issues and some key technologies for practise usage, such as cost/benefit, reliability, survivability and energy efficiency and so on. This thesis studies algorithm design and verification of VNE for cost/benefit and energy efficiency.The thesis focuses at first on the model and performance of virtual nodes splitting for aggregating the diversed physical resources of the substrate physical network. The feasibility of node splitting and a formal expression is presented for generic VNE requests. A mixed integer programming (MIP) algorithm has been proposed by taking both path-splitting and node-splitting into considerations. The algorithm’s performance is evaluated using by a open solver GLPK over a NSFNET topology. The numerical evaluations show that, comparing with the classical VNE without node-splitting, the proposed algorithm is benefit to reduce VNE cost, as well as to increase VN-acceptance-ratio especially when the substrate network resources are to be used up or there is a large amount of virtual network requests.The thesis also analyze the energy consumption in the background of VNE; based on the levels of energy consumption and characteristics of nodes and links, a mathematical model and the corresponding algorithm of energy consumption are constructed. To minimize the energy consumption the energy consumption of virtual network requests, a algorithm named virtual network embedding with rate and energy adaptability (VNE-REA) is introduced on the basis of proportional computing. For purpose of the solution of open software ALEVIN, a binary variable is introduced to represents each rate and energy level. Thus the original nonlinear problem can be transformed into a linear problem. Simulation results show that, VNE-REA can not only improve the acceptance rate of the virtual network requests, but also reduce the energy consumption.