Design And Simulated Implementation Of Virtualized Resource Allocation Algorithms In Fiber-Wireless(FiWi) Access Network

In recent years,the scale of Internet has been expanding with exponentially increased number of network users and endlessly emerging applications,which puts forward higher requirements for the capacity,reliability and flexibility of network.Resulting from the increasing development of optical communication technology and the increasement of backbone network transmission capacity,the access network which covers the "last mile" of communication system becomes the bottleneck to the Quality of Service(QoS)of network.Fiber-Wireless(FiWi)access network,which is the integration of optical access network with large capacity and high reliability and wireless access network with high flexibility and low cost,has catched our eyes.However,the difference of resource management mechanisms between the optical and wireless subnetworks poses a challenge to the seamless integration of FiWi access network,which is hard to satisfy the requirements of novel applications expansion and global resource optimization.To address the issue of network ossification,the network virtualization has been proposed to decouple the Internet Service Provider(ISP)into the service provider and infrastructure provider and achieve the global network resource optimization exploiting the developed hardware and cloud storage technology.Thus,the logically isolated service providing can be obtained by mean of network virtualization.This thesis focuses on the resource allocation mechanism in virtualized FiWi access network.The resource allocation algorithms are proposed by handling the resource allocation pattern challenge of virtualized FiWi access network,the problems of over abstracted virtual network model and insufficient consideration of network resource features that restricts the revenue of InP.These algorithms intend to provide a theoretical guidance and technical reference for the construction and development of virtualized FiWi access network.First,aim to the problem that the resource allocation mechanism in traditional FiWi access network is not applicable in network virtualization scenario where the resource should be preset before the Virtual Network Embedding(VNE),two InP resource allocation algorithms,which are called Wireless mesh network Channel Allocation algorithm based on Breadth first search(WCAB)and FiWi access network Dynamic Bandwidth Allocation algorithm(FDBA)respectively,are proposed to achieve a interference-free scheduling in FiWi access network and minimize the maximum collision domain utilization.Moreover,the virtual network model in virtualized FiWi access network is presented to achieve a more practical description of real world.The node type is defined to express the special Internet-access ability of Optical Line Terminal(OLT).The QoS satisfaction requirements are associated to virtual networks to reflect their different tolerability of QoS satisfaction decline and allow InP accepting a virtual network without 100%requested resource provided.In addition,the Integer Linear Programming(ILP)of VNE problem based on QoS satisfaction is formulated to maximize the InP revenue.And the heuristic algorithm called Virtual Network Embedding algorithm based on QoS satisfaction(VNE-Q)is proposed.Based on the proposed VNE-Q algorithm,the Virtual Network Embedding algorithm based on network Reconfiguration(VNE-R)is proposed with the consideration of the flexibility of resource reconfiguration in virtual and substrate networks.The network congestion type and qualitative cost of different network reconfiguration operations,that is,substrate network bandwidth reallocation,substrate network channel reallocation and virtual network reconfiguration,are considered to improve the virtual network acceptance ratio and InP revenue with low cost.The Visual Studio C++ simulation platform is deployed to evaluate the performance of proposed algorithms and make a comparison with related approaches.The simulation results demonstrate that the proposed VNE-R algorithm preforms apparently better than other approaches in terms of virtual network acceptance ratio and InP revenue.