Virtual Network Embedding Based On Ant Colony Optimization

With the rapid development of network and diversity of applications, the inherent problem exist in Internet is prominent increasingly, such as scalability, control and manage, QOS assurance, green energy and so on. Academia proposed "clean-state" for future network in order to solve these problems fundamentally. Network virtualization is regarded as an important technology to build the next generation of Internet architecture which provide user with a variety of customized services. In network virtualization, network entities include substrate network (SN) and virtual network (VN). Virtual net-work need some nodes and bandwidth. The node of virtual network can be created and deleted expediently. Network virtualization has been used to solve scalability, complex-ity, resource utilization in data center. In addition, in order to realize the sharing of computing resources, separation and aggregation of resource in cloud computing, net-work virtualization has became a key to solve these problems.Virtual network embedding (VNE) that embedding virtual network in substrate network is usually mentioned as resource allocation problem in network virtualization. Virtual network embedding try to reduce the rejection of virtual network requests (VNR) and improving resource revenue on condition that fulfill the resource constraints. Virtual network embedding can be employed to solve the problems like resource constraint, access control, request online and diversity of topology.This paper develops an ant colony optimization algorithm of virtual network em-bedding, ACO-VNE for offline VNR and VNE-CACO for online VNR. ACO-VNE is two-stage embedding algorithm, node mapping first, and then link mapping. The ants secrete and update pheromones in node mapping according to the cost of link mapping. Based on feedback information, the ants move to find good solution through learning from each other. Simulation results suggest that the algorithm can map the virtual net-work with low rejection rate and high revenue of substrate network.