Research On Virtual Network Mapping Technologies

The explosive increase and diversification in user demands and Internet servicespresent many challenges to the traditional Internet structure, and has led to theossification of the Internet. A newly proposed Internet architecture for overcoming theossification of Internet will lead to fundamental changes in the existing Internetstructure, software and hardware of routers. Furthermore, since the Internet isco-provisioned by multiple providers, any change of the Internet architecture needs tobe agreed by these Internet service providers. Due to these factors the prospect ofemergence of a new Internet architecture is very dim. To overcome the ossification onceand for all, network virtualization (NV) as a new paradigm of the future Internet hasbeen proposed.In the network virtualization environment (NVE), multiple VN requests may bemapped onto the same underlying substrate network, and hence transparently share theresources of the same underlying substrate network. The challenging issue encounteredby network virtualization is that how to efficiently map the virtual networks onto theshared substrate network. An efficient VN mapping strategy will increase the utilizationof the substrate network resources and consequently increase the revenues of the InPs,as well as reduce the operating cost of customers. Thus, how to achieve efficient virtualnetwork mapping is a hot topic, researchers have tried to address recently.Chapter1introduces the concepts and related research pertaining to networkvirtualization, as well the issues and state-of-the-art of virtual network mapping.In Chapter2, the author proposes and addresses the issue of how to design an optimalprovisioning scheme for the VN request with elastic resource requirements. Theproblems of optimal provisioning for virtual networks with unsplittable flow (OPVNUF)and optimal provisioning for virtual network with splittable flow (OPVNSF) are bothconsidered in this research. This problem is modeled as a mathematical programmingproblem by using mixed integer programming (MIP), in which both transmission delayand path availability (or failure) are included for accommodating service requests withbetter quality of service (QoS). Based on the formulation of the problem, a Genetic Algorithm (GA) based heuristic algorithm is proposed for solving this problemefficiently.In Chapter3, the author proposes and addresses the issue of how to implement thesurvivable virtual network mapping (SVNM) to withstand large scale substrate networkfailures. This study considers large scale failures classified into a number of regionalfailures. It is only need to preserve resources for each regional failure rather than forindividual failure for efficient using of network resources. Since the allocated resourcesfor a regional failure may also be used for recovering from other failures, there may beredundancies in the allocated resources for all of the regional failures. Therefore, it isnecessary to eliminate the redundant preserved resources to achieve higher resourceefficiency. The author presents the mathmetical optimization model of SVNM problem,as well as devises two kinds of algorithms for solving the SVNM problem efficiently:(i)Lagrangian relaxation based algorithm, LR-SVNM;(ii) Decomposition based heuristicalgorithm, H-SVNM.In Chapter4, the author investigates the online stochastic VN mapping (StoVNM)problem, in which each bandwidth demand x follows a normal distribution. Such aprobabilistic characterization can better represent the uncertainty of the futurebandwidth usage. In the VN edge mapping process, the consumed bandwidth resourcesare reduced efficiently by stochastic multiplexing. The author formulates the model forthe StoVNM problem by using mixed integer linear programming, as well as proposes asliding window approach based heuristic algorithm w-StoVNM for solving this model.In Chapter5, the author proposes and studies the problem of optimal provisioning forevolving virtual network requests. When a VN request evolves, it is necessary toreconfigure the existing virtual network in order to adapt to its changing requirementsand remap it intelligently so that the resources of the underlying substrate network areused efficiently. This research considers four different scenarios of evolving VNs, andformulates the problem as a mathematical optimization problem, and proposes aheuristic algorithm (Dynamic Virtual Network Mapping Algorithm, DVNMA).In Chapter6, the author studies the power-efficient online mapping technology forvirtual network requests. To achieve this, the proposed solution consolidates VN nodesonto fewer servers and turns the “unnecessary” servers off, as well grooms the WDMnetwork traffic into fewer wavelengths and shuts down the “unnecessary” wavelength on fiber link for saving power consumption. The author models the power-efficientprovisioning for virtual network requests as a mathematical programming problem anddesigns an efficient heuristic algorithm for solving this problem.For validating and evaluating the correctness and effectiveness of the algorithmsproposed in this paper, the author conducts extensive simulation experiments. Chapter7concludes all of the research in this dissertation and presents directions for futureresearch.