Dynamic Resource Allocation Based on Q-learning for VNE in Fiber-Wireless (FiWi) Access Network.

As a promising candidate architecture for future access network, Fiber-Wireless (FiWi) access network is also suffering from the bottleneck of resource allocation and optimization due to the complexity and diversity of traffic demands. Although the emerging network virtualization technology has provided a feasible way for FiWi to break the bottleneck, previous works ignored the dynamic characteristics of resource demand of virtual networks which resulted in the resource wasting. In this paper, a dynamic resource management mechanism based on Q-learning algorithm in FiWi access network is proposed. Each physical node and link is equipped with an agent whose responsibility is to detect the delay or packet loss rate of the virtual network embedding periodically and adjust the corresponding resource demand to meet the specific QoS requirement of the virtual networks. Simulation results demonstrate that the proposed dynamic resource management mechanism is able to improve the utilization of network resources on the premise of meeting the QoS requirement of virtual networks.