Abstract
In this thesis, we study the concept of IaaS. Ifrastructure Providers (InP) management data center's resources through virtualization technology. InPs manage areas such as power consumption, storage, network architecture. Although the virtual networks provide hardware resources to tackle the performance and network security issues, InPs have to face another challenge that is virtual network embedding problem. The virtual network embedding problem is how efficient and flexible allocated to the virtual network into hardware resources. However, the virtual network embedding problems in the data center is developed only in recent years. We propose a heuristic algorithm for mapping virtual network of random topology into a form of star topology that is suitable tree structure of data center. We design a scheme to quantify virtual networks and admission control when virtual network request is coming. We exploit the advantages of the traditional bin packing problem algorithm FFD and BFD to make the virtual network dividing into several clusters, and mapping into the substrate network to minimize link cost. Finally, we use the min cost max flow algorithm to map the virtual link into physical links. Through simulation experiments we verified our idea in terms of objective function, e.g. cost, revenue-cost ratio. We find our method indeed better than traditional bin packing algorithms to better performance.