| The explosive growth of the Internet traffic and the increasingly diver-sified and personalized network application pose great challenges on current networks.In addition,the problems faced by current networks have become in-creasingly prominent,such as poor scalability,lack of QoS,and security.These issues and challenges have driven the research of future network architectures and key technologies.In this context,it has become a trend to integrate net-works,storage,and computing into one system.The unified management of these underlying resources will significantly improve the application and sys-tem performance.The research work of this thesis includes three aspects:the design of in-tergraded framework,multi-dimensional resource capacity allocation and joint multi-dimensional resource scheduling.The main contributions are as follows:1.Integrated networking,storage,and computing designExisting software-defined networking technologies and virtualization tech-nologies can realize the programmable networking,but the research on con-trolling and scheduling computing and storage resources are still independent.This thesis presents a integrated framework called software-defined network-ing,caching and computing(SD-NCC),which can dynamically schedule net-working,storage and computing resources to meet the needs of different appli-cations.2.Energy efficiency capacity allocation of multi-dimensional resourcesIn order to comply with the development trend of energy efficiency com-munications of Internet,the rational capacity allocation of multi-dimensional resources can greatly promote the development of the new SD-NCC frame-work.By considering energy consumption,this thesis studies the optimal ca-pacity allocation strategy of multi-dimensional resources to minimize this joint objective function.The thesis derives the optimal solution through derivation analysis.3.Joint allocation of multi-dimensional resourcesThe new SD-NCC framework achieves excellent system performance through jointly scheduling of networking,caching and computing resources.This dis-sertation studies the joint allocation strategy of multi-dimensional resources from two perspectives:a centralized way and a distributed way.1)Research on the centralized joint allocation of multi-dimensional re-sources based on load balancing:This dissertation studies the joint resource allocation by considering two factors(minimization of network usage and load balancing).By considering the network state information(NSI)obtained by the controller of the control plane,the research of joint resource allocation prob-lem is conducted for the perfect NSI and unperfect NSI.Under perfect NSI,two solutions are proposed:a problem decomposition method and a problem transformation method.Under unperfect NSI,an adaptive resource allocation management algorithm under time-varying network requirements is proposed based on discrete random approximation(DSA).2)Research on distributed allocation strategy of multi-dimensional re-source:In order to reduce the computational complexity and signaling overhead of centralized algorithm,this dissertation studies the distributed allocation strat-egy of multi-dimensional resources.An ADMM-based algorithm is proposed,which can greatly reduce the signaling overhead between the control plane and the data plane.Finally,the thesis summarizes the whole research work,and discusses the future research direction. |
