Research On Virtual Subnet Embedding Algorithms For Wireless Mesh Networks Based On Hopfield Neural Network

Wireless mesh network(WMN) is a new kind of wireless network technology that is completely different from the tradional wireless network. It has a very broad application prospects. WMN has lots of obvious advantages including low cost, conveni ent deployment, wide coverage, high throughput,etc. However, several challenging problems emerge in actual deployments, such as not guaranteed bandwidth for a user, relatively high packet loss, and easy to link disruption. Wireless mesh network virtualization is adopted to overcome these problems.We mainly studied the virtualization in wireless mesh network in this thesis. Firstly, we presented the related work of WMN and virtualization,and also introduced the necessity and application about virtualization in WMN. Secondly, we introduced the classification framework for scheduling algorithms in WMN. Based on the virtual environment, the object that is scheduled is virtual subnet, instead of users and links. We summarized the current algorithms of virtual network embedding(VNE) in WMN. At last, the allocation of resource and the problem of VNE was solved by the algorithm of Hopfield neural network.Virtual network embedding is one of the most challenging problem in network virtualization. The problem of VNE is a joint optimization that includes resource scheduling, the mapping of virtual nodes and virtual links. We divided the problem of VNE into two parts, the optimization of resource scheduling and the mapping of subnet, since it is too complicated and is still NP-hard. In this paper, the OFDMA mesh system was adopted and subchannels was scheduled. Unlinke wire networks, in terms of the mapping of virtual nodes in virtual networks, the embedding strategy of virtual nodes is determined by their positions. For the two combinatorial optimization problem, subchannels assignment and virtual links mapping, we adopted Hopfield neural network that has the advantages of parallel computation and fast convergence to solve this problem.At last, the results of simulation showed that the algorithm of Hopfield neural networkis effective for solving the problem of subchannels assignment and the mapping of virtual links. And the multi-path mapping can obviously increase the virtual subnet request acceptance ratio and the utilization of the substrate network resource. In the end, for the virtual subnet that arrived simultaneously, we simulated the mapping depending on their priority and given the simulation results..